Working Paper
Updating the Diamond and Morlino Framework for the Algorithmic Age
Diamond and Morlino's eight dimensions of democratic quality remain one of the most influential frameworks for evaluating democratic performance. Developed in 2004, however, they presuppose an environment in which democratic agency is formed through institutions and communicative processes that are broadly transparent and contestable. This paper argues that the widespread deployment of algorithmic systems has fundamentally altered those conditions. Contemporary algorithmic architectures shape political information, structure public reasoning and increasingly mediate the exercise of state power before citizens participate in democratic life, creating forms of influence and opacity that the existing framework cannot adequately capture.
Drawing on Habermas, Zuboff, Dean, Danaher, Pasquale, Gowder, Klonick, Srivastava and, normatively, Forst's concept of noumenal power, my paper develops an architectural account of democratic agency. I argue that algorithmic systems exercise constitutive power over the conditions under which citizens form political judgements and encounter administrative authority, while remaining insufficiently visible, contestable and democratically accountable. Examining the dimensions of participation, responsiveness and the rule of law demonstrates that these effects do not expose weaknesses within Diamond and Morlino's framework itself; rather, they reveal a structural assumption shared by all eight dimensions: that the environment in which democratic activity occurs is capable of sustaining meaningful agency.
My paper proposes a ninth dimension of democratic quality concerned with the democratic governance of algorithmic architectures. Rather than replacing the existing framework, this additional dimension complements it by evaluating whether the informational and administrative conditions that make democratic agency possible remain transparent, contestable and subject to public justification. In doing so, the paper extends democratic quality theory to address the challenges posed by platform governance and algorithmic decision-making in contemporary democracies.
Why does Diamond and Morlino’s 2004 framework require updating? In April 2026, over 300,000 individuals signed two petitions calling on the UK government to terminate its contracts with Palantir Technologies (38 Degrees, 2026a; 38 Degrees, 2026b). These contracts, covering the NHS, MoD and multiple police forces, were awarded through established parliamentary and administrative processes. With a combined value exceeding £600 million and strong indications of further expansion (NHS England, 2023a; Ministry of Defence, 2025; HM Government, 2025), such contract awarding would ordinarily pass as routine business, allowing governance to proceed without democratic challenge. In this case, however, the contracts and the backlash they have generated provide a useful starting point for examining how democratic accountability operates in relation to technologically complex and morally contested systems.
What distinguishes these contracts is not simply their scale; they have the potential to impact all individuals who have data recorded in the UK regardless of nationality. The openly stated views of Palantir’s founders, the company’s deep involvement in global security and military logistics and ongoing concerns around data sovereignty, particularly in relation to the US CLOUD Act, combine to produce a form of governance that sits uneasily within democratic expectations. This tension grew in April 2026. Palantir published a twenty-two point manifesto on X, drawn from Karp and Zamiska (2025), setting out a clear, technology-driven political position (Palantir Technologies, 2026). Its dismissal of concerns around AI weaponry as “theatrical” (point 5), reframes issues of accountability as secondary to technological inevitability. Similarly, classifying equality as “dogma” (point 21), positions cultural identity within a grading system, rejecting some cultures as less valuable than others.
Palantir represents a particularly visible instance of this transformation. It is not, however, exceptional. In the UK, the Public Law Project’s TAG register identifies 55 active algorithmic systems, most of which operate with limited transparency and restrict meaningful scrutiny (Public Law Project, 2024b). The Department for Work and Pensions’ Universal Credit Advances Model illustrates the problem well. As a machine learning system used to assess fraud risk in advance payment requests, it operates under substantial opacity. Its architecture, specifications and operational data are withheld under Freedom of Information exemptions. Claimants subject to its decisions are therefore unable to assess the basis on which those decisions are made (Department for Work and Pensions, 2025b).
Where scrutiny emerges, it is typically reactive and partial. The DWP’s fairness analysis, published in July 2025 following external pressure and covering April 2024 to March 2025, found that older claimants and non-UK nationals were disproportionately over-referred for review. These findings were later confirmed by the National Audit Office (Department for Work and Pensions, 2025a; National Audit Office, 2025). Similar dynamics appear beyond the UK. In the Netherlands, the SyRI welfare fraud detection system was ruled unlawful by the District Court of The Hague in 2020 for violating the right to privacy under the European Convention on Human Rights. This was the first case in which a state algorithmic system was invalidated on such grounds (van Bekkum and Zuiderveen Borgesius, 2021, p. 324).
Together, these cases point to a transformation in how political power operates. Existing measures of democratic quality do not adequately capture this shift because they assume forms of decision-making that remain visible, attributable and challengeable. Where governance depends on opaque and privately developed systems, these assumptions no longer hold. I argue that an additional element is required, one capable of holding to account the implications of algorithmic dependence, opacity and the erosion of contestability within decision-making itself.
Diamond and Morlino’s eight dimensions of democracy (Diamond and Morlino, 2004) frame how we can judge the strength of a democracy. The dimensions work well, serving to frame what are normative aspects of democratic life. They are as equally valid in 2026 as they were in 2004. However, what they do not do is account for the presence of technology, particularly the use of algorithmic decision-making systems. They embed assumptions, prioritise particular forms of knowledge and, as in the case of Palantir, are increasingly accompanied by explicit political visions that sit outside democratic scrutiny. The framework's limitations in this respect are not a flaw in its design. They reflect the conditions under which it was constructed. Diamond and Morlino's dimensions rest on an assumption that was reasonable in 2004: that the environment within which democratic activity takes place is broadly neutral. Citizens are treated as political subjects who form preferences, exchange reasons and exercise their right to hold their government accountable. The eight dimensions measure how well democracy performs within that environment. What they do not measure is whether the environment is now capable of sustaining the conditions those dimensions require; it could not.
The 2026 digital environment plays a constitutive role, shaping what political information citizens engage with, how it is weighted and delivered, how engagement is made available and how state power is increasingly exercised. This happens before individuals deploy any of the capacities the dimensions are designed to reflect. This constitutive role is most precisely understood through Forst's concept of noumenal power: "to have and to exercise power means to be able—in different degrees— to influence, use, determine, occupy, or even seal off the space of reasons for others"(Forst, 2015, p. 116). Algorithmic systems exercise precisely this kind of power, shaping the normative space of democratic agency before it is deployed.
Vertical accountability, as Diamond and Morlino conceive it, describes the relationship between citizens and the state, the ways in which citizens form political judgements, make demands of power, hold government to account and through which the state exercises power over individuals in ways that remain visible, attributable and challengeable. Algorithmic systems have corrupted this accountability process at both ends. They degrade the upward flow of citizen agency and the downward exercise of state power in ways that the existing dimensions of democratic quality cannot detect. Understanding how the degradation occurs requires examining each direction in turn.
The first direction is upward. Vertical accountability depends on citizens being able to form coherent political judgements, exchange reasons and exercise pressure on those who govern them. The algorithmic architectures of online platforms, search engines, content aggregators, recommendation systems and social media, shape the informational environment through which this upward flow of accountability operates. This is not just a social media problem. It is a platform architecture problem. Google's search ranking determines what political information is findable. YouTube's recommendation system shapes what political content is consumed and in what order. News aggregators algorithmically curate what counts as relevant journalism. Critically, these systems do not deliver the same environment to all citizens in the way that, for example, newspapers or magazines do. Decisions made by and about individuals include their search history, location, demographic characteristics and inferred preferences. These are some of the factors which determine what each person sees. This produces divergent and personalised information environments that fragment the shared epistemic foundation on which democratic activity depends. What one citizen understands to be politically true, urgent or legitimate may bear little resemblance to what another encounters, not because of differences in their reasoning, but because of differences in what their algorithmic environment has made available to them. Pariser identifies this dynamic as the filter bubble, the personalised informational enclosure produced by algorithmic curation (Pariser, 2011, pp. 9–10), while Sunstein demonstrates that such enclosures produce echo chambers that reinforce existing beliefs, systematically excluding the kind of exposure to alternative perspectives that democratic deliberation requires (Sunstein, 2017, pp. 9–11). Across all of these systems, what is visible, credible and shareable in the public sphere is both shaped and individualised before citizens exercise any of the capacities on which vertical accountability depends. As Alnemr demonstrates, these systems promote political polarisation, undermine the epistemic quality of public deliberation and compromise the conditions of rational political agency before citizens enter the public sphere at all (Alnemr, 2025, pp. 4–5). Habermas argues that the platform revolution represents not an expansion of existing media but a structural break (caesura) comparable in significance to the introduction of printing, one in which algorithm-steered platforms fragment the shared epistemic basis of the public sphere into self- enclosed echo chambers that acquire the status of competing publics. This fundamentally alters the conditions under which democratic opinion and will formation can operate (Habermas, 2022, pp. 158, 162). The result is that vertical accountability is weakened at its foundation. Citizens hold governments to account on the basis of preferences and beliefs that have been algorithmically shaped and individually differentiated before the accountability process begins.
The second direction is downward and more opaque. Vertical accountability requires that the state's exercise of power over individuals remains visible, attributable and challengeable. Where it does not, citizens cannot contest decisions that affect them, representatives cannot examine the actions of government and the democratic relationship between state and citizen is hollowed out from within. Algorithmic systems embedded within state institutions have created this condition. Ulbricht and Egbert demonstrate the problem clearly in their study of Palantir's deployment in German federal policing. The modular and proprietary character of such platforms makes it exceptionally difficult for parliamentary opposition, civil society and affected citizens to assess either the scope of data use or the reasoning that produces decisions affecting them: the ‘what’ and the ‘why’. They conclude that platform regulation in the public sector “is not only about technology-specific regulation but also about general mechanisms of democratic control, such as the separation of power, public transparency, and civil rights” (Ulbricht and Egbert, 2024, p. 12). The MoD's own procurement documents acknowledge that its data analytics capabilities have become so dependent on Palantir architecture that changing supplier would cause disproportionate technical difficulties (Ministry of Defence, 2025). This is institutional lock-in stated openly in an official government document, creating a form of algorithmic dependence that places the exercise of state power beyond the reach of democratic accountability by design rather than by accident. The DWP's Universal Credit Advances Model illustrates the same condition at the level of individual citizens. The system withholds its model architecture, specifications and operational data under Freedom of Information exemptions, meaning that claimants subject to its decisions cannot scrutinise the basis on which those decisions are made (Department for Work and Pensions, 2025b). When the DWP's own fairness analysis found that older claimants and non-UK nationals were being disproportionately over-referred for review, the finding was not proactively published — it emerged only after sustained external pressure (Department for Work and Pensions, 2025a; National Audit Office, 2025). The system exercises power downward over individuals while remaining opaque to those individuals, to civil society and to the elected representatives who are formally responsible for it. Beyond the UK, the District Court of The Hague's ruling against the SyRI system established that this is not merely a failure of administrative practice but a violation of fundamental rights, the first judicial recognition that state algorithmic systems can breach the right to privacy under the European Convention on Human Rights (van Bekkum and Zuiderveen Borgesius, 2021, p. 324). What these cases share is not a problem of transparency. It is a specific form of opacity, one that is structural, legally protected and increasingly treated as a normal feature of public administration rather than an exception to it. The normalisation has embedded itself into democracy and is being used to justify its own use. The individual subject to these systems encounters state power removed from accountability processes. They cannot identify the assumptions encoded in the system, contest the reasoning that produced the decision affecting them, and, as the Palantir case illustrates, may find that the company whose system governs their relationship with the state holds an explicit political vision that has never been subject to democratic deliberation. Forst's concept of noumenal power captures this condition precisely. These are systems that seal off the space of reasons for those subject to them, structuring what citizens can demand, contest and justify in their relationship with the state before any formal accountability mechanism is engaged (Forst, 2015, p. 116).
Both directions corrupt the dimension of participation in ways that are formally invisible. Participation can register as robust—citizens vote, petition, campaign and engage—while the conditions of participatory agency have been hollowed out simultaneously at both levels. On the upward direction, political preferences are shaped before citizens engage with one another in the public sphere. The platform environment pre-structures what citizens understand themselves to want, believe and oppose before they participate in any accountability process. On the downward direction, the state's exercise of power over citizens is increasingly mediated by algorithmic systems that participation itself cannot reach. A citizen may petition against the use of machine learning in welfare decisions, as 300,000+ have done in relation to Palantir, while the systems that govern their relationship with the state continue to operate behind FoI exemptions and proprietary architectures that place them beyond public reach. This is not a failure of participation in the formal sense. Citizens retain the right to vote, to organise, to petition and to challenge. What has changed is the environment in which those rights are exercised. On one side, the informational conditions of participation have been algorithmically pre-structured and individually differentiated before citizens engage. On the other, the decisions that participation is supposed to influence are increasingly produced by systems whose logic participation cannot enter. The Diamond and Morlino framework provides guidance for measuring the activity of participation. It has no mechanism for assessing whether the conditions that make participation meaningful—shared information, contestable reasoning, visible and attributable state power—remain intact. The eight dimensions remain capable of measuring the outputs of democratic life. This is tested through a range of empirical rubrics such as V-Dem, Freedom House's Freedom in the World index and the Economist Intelligence Unit's Democracy Index. What they cannot assess is whether the prior conditions that give those outputs democratic substance are being systematically eroded. A democracy can score well across all eight dimensions while the architectural foundations of meaningful agency are quietly hollowed out. Citizens can participate formally while being unable to contest the systems that shape what they know, what they believe and what the state does to them. Accountability appears to function while power is being exercised by the algorithmic layer. This is the gap. It is structural rather than empirical, and it cannot be addressed by refining the existing dimensions. The dimensions ask how well democracy performs. The gap concerns whether the conditions under which democratic performance is possible remain in place. Closing it requires not a revision of Diamond and Morlino's framework, but an addition to it, a ninth dimension directed at the conditions under which democratic agency becomes possible: the visibility, contestability and democratic answerability of the algorithmic systems. The following sections develop the theoretical foundations for that dimension, drawing on Forst's account of noumenal power and the distinction between the upward and downward directions of vertical accountability, before setting out its content, indicators and regulatory implications.
The argument proceeds in three steps. First, drawing on Habermas (2022) and Zuboff (2019), I establish what architectural neutrality requires of a communicative environment and how its absence compromises both the horizontal conditions of political communication and the vertical conditions of democratic governance. Second, drawing on Dean (2005), Danaher (2016) and Pasquale (2019), I identify the consequences of that loss: why citizens cannot recognise the structural erosion of their agency and why existing governance frameworks cannot reach the architectural level at which that erosion occurs. Third, drawing on Gowder (2023), Klonick (2018) and Srivastava (2021), I establish that platforms operating at scale exercise a form of power similar to sovereign authority, approached from political theory, legal-constitutional analysis and international relations respectively. Forst (2017, 2024) provides the normative ground for why that analogy generates democratic obligations. Democratic legitimacy, on Forst's account, rests on a prior obligation of justification owed by those who exercise power to those subject to it. Platforms have acquired the reach of that power without accepting its justificatory obligations. It is the gap between the power platforms exercise and the accountability they accept that the ninth dimension is designed to capture.
Democracy depends on informed public opinion. This allows for participation and accountability to happen. In this respect, Habermas's account of the ‘public sphere’ provides a useful starting point for understanding the democratic threat posed by algorithmic systems. He locates the public sphere within the "functionally differentiated structure of modern societies between civil society and the political system" (Habermas, 2022, p. 146). Civil society here is understood as the everyday practices of public discussion, association and opinion formation among citizens. The political system is understood as the institutional setting in which binding collective decisions are made, a core condition of democratic government. The public sphere operates between civil society and the political system by organising public opinion and transmitting it toward political authority. This provides a normative baseline. For the public sphere to perform this function, it must be capable of operating without distortion, carrying the authentic preferences and demands of citizens toward the institutions that are accountable to them. What is constitutive of the public sphere, on Habermas's account, is not the form participation takes but "the topics that deserve shared interest and the respective professionally examined form and rationality of the contributions" that make mutual understanding possible (Habermas, 2022, p. 165). In Diamond and Morlino's terms, this is the prior condition of vertical accountability. They define accountability as "the obligation of elected political leaders to answer for their political decisions when asked by voters or constitutional bodies" and note that it requires "fairly robust levels of voter interest, information, and turnout" (Diamond and Morlino, 2004, p. 25). Citizens can only hold governments to account if the opinions and preferences they bring to that process are authentically their own. Where the public sphere is distorted, vertical accountability is compromised before it begins. Diamond and Morlino's framework measures accountability at the institutional level, asking if elections are free and fair, do leaders answer for their decisions and if citizens can remove governments from office (Diamond and Morlino, 2004, p. 25). These are vital measures. But they presuppose that the opinions and preferences citizens bring to those institutional processes are authentically their own. Where algorithmic systems shape what citizens encounter, curate their engagement and determine what they understand themselves to want before any institutional interaction takes place, that presupposition fails. It is this prior architectural level, the conditions under which democratic agency itself is formed, that Diamond and Morlino's framework does not yet reach.
Habermas develops this argument directly. He argues that algorithm-controlled platforms are "far from neutral" in their operation, actively distorting the deliberative conditions on which democratic legitimacy depends (Habermas, 2022, p. 163). Part of what makes this distortion so significant is that platforms are not simply a new form of media. Habermas is explicit on this point: the new media "are not 'media' in the established sense" because, unlike traditional press, radio or television, they "neither produce, nor edit nor select" communicative content (Habermas, 2022, p. 159). Instead they act as intermediaries "without responsibility", establishing connections and intensifying discourses with "unpredictable contents" while profoundly altering "the character of public communication itself" (Habermas, 2022, p. 159). This distinction matters for the argument advanced in this paper. The democratic threat identified here is not solely a social media problem, though social media is part of it. It is an algorithmic problem. What is at stake is not the medium through which communication occurs but the algorithmic logic that increasingly governs what citizens encounter, how institutions process demands and how decisions about individuals are made. Whether that logic operates through a social media feed, a welfare eligibility system, a policing tool or a border control algorithm, the democratic stakes are the same. The medium is visible. The algorithm is not. Where platforms once promised to extend and democratise the public sphere, giving voice to citizens previously excluded from editorial media, that promise has been overtaken by a commercial logic that fragments public communication into self-enclosed echo chambers, rewarding engagement over deliberation and sensation over reasoned exchange (Habermas, 2022, pp. 159-160). The result is a structural weakening of the conditions under which public opinion can form in ways that are democratically meaningful. What citizens experience as political participation increasingly reflects an algorithmically curated environment rather than the kind of open rational-critical exchange the public sphere requires. This structural condition is not simply contextual. It is located in the architecture of platforms rather than in the behaviour of individual users or the content they encounter. This matters because it shifts the level at which democratic analysis needs to operate. If the problem were behavioural or contextual, existing and evolving regulatory mechanisms could resolve this. Because it is architectural, they cannot. However, Habermas's account has a significant limitation for the purposes of this investigation. His primary concern is with how algorithmic fragmentation disrupts the formation of public opinion and weakens its transmission toward political authority. In Diamond and Morlino's terms, this is a problem of vertical accountability. On their account, citizens cannot effectively hold governments responsible if the opinion formation process has been structurally distorted before it reaches institutional processes (Diamond and Morlino, 2004, p. 25). The disruption of citizen-to-citizen communication that Habermas identifies is part of this problem, but it is not the whole of it. What Habermas does not fully address is a second and deeper facet of the architectural problem. Where the first operates within the public sphere, distorting the conditions of citizen communication and opinion formation, the second operates entirely outside it. This is the executive or administrative face of algorithmic power. It encompasses the procurement and deployment of algorithmic systems by public authorities to make or inform decisions about citizens in domains such as welfare, policing and border control. These systems do not distort participation in the public sphere. Prior to the use of digital algorithmic tools, they did not form part of the public sphere. The systems bypass the public sphere entirely, shaping decisions about individuals before any democratic process engages. It is to this dimension, and to the theoretical tools necessary to diagnose it, that the investigation now turns.
Zuboff's account of surveillance capitalism clarifies the commercial logic driving both sides of the architectural problem. In The Age of Surveillance Capitalism (2019), she identifies a fundamental transformation in the operational logic of digital platforms. What platforms present as open participatory environments are, architecturally, systems designed to extract behavioural data, convert it into predictive products and deploy those products to modify future conduct. It takes the now and creates the then. Platforms are not neutral carriers of human interaction. They are active participants in shaping it, positioned toward commercially determined ends that are structurally hidden from the scrutiny of those subject to them. Zuboff terms this form of power "instrumentarianism", a power that "knows and shapes human behaviour toward others' ends" operating not through direct coercion but through "the automated medium of an increasingly ubiquitous computational architecture of 'smart' networked devices, things and spaces" (Zuboff, 2019, p. 8). This formulation is significant for two reasons. First, it provides an explanation of the logic beneath Habermas's observation that platforms operate "far from neutral" and act as intermediaries "without responsibility"—the distortion of the public sphere is not incidental to platform design but structurally necessary to it. Second, and crucially for the argument advanced here, Zuboff's own language reaches beyond consumer-facing platforms. Smart networked devices, things and spaces: the architecture of welfare eligibility systems, facial recognition tools and border control algorithms. More significantly, it is the architecture within which political reasoning increasingly occurs. The implications of that observation for democratic legitimacy will be developed in Stage 3 of this argument. For the purposes of this investigation, what is significant is not simply the commercial logic Zuboff identifies, but its impact on democracy. Where the informational environment through which citizens encounter political life is created by systems organised for engagement and behavioural modification rather than deliberative exchange, the conditions of the public sphere are compromised before political activity begins1. Here Zuboff's diagnosis partners with Habermas's normative baseline. What appears as open participation moves, architecturally, from what is actually delivered. The gap between appearance and operation that Zuboff identifies is not a consumer protection problem which could be remedied with legislation. It is a structural condition of democratic life.
Zuboff's account addresses the first facet of the architectural problem more fully than the second. Her analysis remains primarily oriented toward consumer-facing platforms, the visible, user-facing architecture of social media and search engines. What it does not fully address is the second facet: the deployment of algorithmic systems by public authorities to make or inform decisions about citizens outside the public sphere. The extractive logic she identifies is not confined to the relationship between platforms and their users. The UK's deployment of algorithmic tools by the DWP to assess benefit eligibility and the use of live facial recognition by UK police forces illustrate how the same structural conditions—opacity, insulation from contestation and the prior structuring of outcomes— reappear in the governance context. In both cases, decisions about individuals are shaped by systems whose logic is proprietary and resistant to conventional challenge (for a fuller treatment of these and related cases, see Stage 1). The difference from the consumer-facing context is significant: the power being exercised is not commercial but administrative. Its consequences for individuals as political agents are correspondingly more direct and immediate. This application of Zuboff's logic is not merely speculative. Regulatory and legal developments across the UK and Europe confirm that the democratic implications of algorithmic deployment are increasingly recognised and contested.2 The deployment of algorithmic decision-making systems by public authorities has generated a growing body of legal and regulatory scrutiny, precisely because the accountability frameworks designed for human administrative decision-making do not map cleanly onto systems whose logic is locked away from conventional forms of interrogation. What Zuboff identifies as a feature of surveillance capitalism, the structural insulation of algorithmic logic from meaningful challenge, reappears in the governance context as a feature of algorithmic administration. Across both the commercial and administrative domains, the consequence is the same: individuals are subject to systems that shape their political environment and their administrative reality without being able to challenge the logic by which those systems operate.
A critical dimension of this architectural problem is identified by Dean. In ‘Communicative Capitalism: Circulation and the Foreclosure of Politics’ (2005), she argues that contemporary capitalism has transformed political communication into what she terms ‘communicative capitalism’, a condition in which the circulation of messages has become an end in itself rather than a means to political action. Citizens engage continuously, but that activity does not accumulate into political will or meaningful democratic change. Dean terms this the ‘fantasy of participation’: the feeling of engagement substitutes for its democratic substance (Dean, 2005, pp. 60-61). Crucially, this is not a claim about deception or manipulation. Citizens are active and engaging. The problem is structural. The architecture of communicative capitalism captures political energy and recirculates it without directing it toward democratic ends. Core to this is the purpose of the algorithm. Its success is measured not by democratic quality but by commercial performance — hits, length of engagement, return visits. Democratic activity is not a KPI. If democratic engagement occurs, it is a byproduct of a system designed for something else entirely. Where political content drives engagement the algorithm will amplify it not because it serves democratic ends but because it serves commercial ones. Divisive, emotionally charged political content drives engagement. The architecture does not distinguish between deliberation and outrage. Both register as engagement.
The empirical picture of internet use in 2025 illustrates the scale of this condition. More than half of all internet users globally report finding information and keeping up to date with news and events as primary reasons for using the internet, with 62.8% and 55% respectively citing these motivations (Kemp, 2025, p. 65). This suggests an active and informed citizenry. However, the infrastructure through which that engagement occurs is dominated by platforms that are among the most legally and regulatorily contested. Included in the top ten of the most visited websites globally are those most algorithmically governed platforms, each optimised for engagement, retention and behavioural modification (Kemp, 2025, p. 109). The sites are Google, YouTube, Facebook, Instagram and X. Citizens are informing themselves and following events, but doing so within environments whose architectural logic has been established, through Habermas and Zuboff, to not operate neutrally. Dean's analysis explains why this matters: not to show that citizens are disengaged, but to show why sustained, high levels of engagement are not translating into meaningful democratic agency. Dean's argument does something analytically important that Zuboff's does not. Where Zuboff identifies what is happening at the architectural level, Dean explains why it remains invisible to those subject to it. The experience of navigating the digital environment is indistinguishable from meaningful democratic participation. Citizens are genuinely doing things: contributing to a debate, signing a petition or reading a report. What they cannot perceive, from within that experience, is the degree to which the architecture has pre-structured what they encounter, how they engage and what kinds of political demands their activity is capable of generating.3 This has a direct consequence for the accountability argument. If citizens cannot perceive the structural conditions shaping their agency, they cannot identify what requires justification or to whom a justificatory demand should be addressed. The accountability problem is not simply a failure of regulators or legislators to act. Increasingly they do. Citizens cannot demand accountability for conditions they cannot perceive. The problem is epistemic: citizens do not know what to challenge or how to challenge it. The architecture that shapes democratic agency is the same architecture that makes that shaping invisible. Zuboff identifies the opacity of the platform. Dean identifies the opacity of the experience. The result is a compound problem. The justificatory gap is structurally locked from the democratic challenge that might otherwise close it.
Dean's account is primarily concerned with citizen communication and political expression, which is the face of the architectural problem that operates within the public sphere. Like Zuboff, however, it addresses the second facet only partially: the conditions under which political demands reach institutions that are themselves dependent on algorithmic systems operating beyond the reach of democratic challenge. A citizen may experience themselves as politically engaged, yet the degree to which they can perceive or contest the conditions shaping that experience is structurally limited. They will be unaware that decisions shaping their access to welfare, policing actions or their status at a border have already been made by systems whose logic they cannot inspect or contest. The illusion of participation operates on both sides of the architectural problem: in the public sphere and in the administrative domain where citizens are subjects rather than participants. Dean’s identification of why citizens cannot perceive the erosion of their democratic agency opens the question of what conditions enable this to happen, the structural conditions that undermine democratic activity.
Danaher provides an approach to framing the structural condition that produces it. In 'The Threat of Algocracy: Reality, Resistance and Accommodation' (2016), he defines algocracy as a governance system "in which algorithms are used to collect, collate and organise the data upon which decisions are typically made", structuring and constraining "the ways in which humans within those systems interact with one another, the relevant data and the broader community affected by those systems"(Danaher, 2016, p. 247). The threat he identifies is located in what he calls the ‘opacity concern’ and the “growing use of algocratic systems" undermining “political authority and legitimacy” (Danaher, 2016, p. 249). The deficit produced by algorithmic governance is not a problem that data protection law or privacy regulation can resolve. It operates at the level of political authority and legitimacy, as Danaher notes, prior to and beyond the reach of those frameworks. The democratic weakening of this condition is captured in Danaher's argument. For him, legitimate decision-making procedures require participation in and comprehension of those procedures by the citizens they affect. Increasing reliance on algocratic systems limits both. The more algocratic a system is, the less legitimate it becomes, producing what Danaher terms the 'human-out-of-the-loop' condition (Danaher, 2016, p. 254).4 It is a structural condition that existing democratic mechanisms were not designed to address. Those mechanisms were designed for a world in which decision-makers are human, closer to Danaher's human-in-the-loop model, and can be identified, questioned and held responsible. Contemporary algocratic systems resist this presupposition at every point. In Diamond and Morlino's terms, the implications reach across multiple dimensions simultaneously. Participation is compromised where citizens engage within environments whose architectural logic pre-structures what they encounter. Responsiveness is compromised where governments respond to preferences already shaped by systems optimised for engagement rather than deliberation. The rule of law is compromised where legal frameworks designed for human decision-makers cannot reach the algorithmic conditions that precede their application. Each dimension registers a symptom. None reaches the structural condition that produces it.4
Writing in 2016, Danaher provided some prescience. We may, he suggests, "be on the cusp of creating a governance system which severely constrains and limits the opportunities for human engagement, without any readily available solution"(Danaher, 2016, p. 266). His concern for the future is instructive for our current position, in 2026, and for this argument in two key respects. First, the problem cannot be resolved by incremental adjustments to existing mechanisms. The structural condition is not open to procedural fixes. Second, Danaher's account remains primarily concerned with governance systems in the conventional sense: the bureaucratic, legislative and legal processes increasingly reliant on algorithmic assistance. These are valuable contributions to democracy discussions. What is not fully addressed is the relationship between algocratic governance and the commercial platforms that shape the conditions of democratic agency in the public sphere. The opacity concern Danaher identifies operates on both facets of the architectural problem, but the legal and institutional frameworks he examines as possible responses are oriented primarily toward the administrative facet and, as he concludes, are unlikely to be sufficient.5 Where Danaher diagnoses the structural condition of algocratic governance, Pasquale identifies the legal dimension and intractability of interpretation of outputs. In doing so, he identifies what accountability would require and why it remains structurally out of reach in the absence of attribution to particular persons.
In 'A Rule of Persons, Not Machines: The Limits of Legal Automation' (2019), Pasquale demonstrates that the problem emerges as soon as algorithmic systems move beyond the trivial. Even in a relatively simple automated contract—an airline and insurer automating payments linked to oil prices—the moment something unexpected occurs, for example a bank payment failure, the system has no capacity for the contextual judgment that fairness requires. As Pasquale observes, "it is a far more formidable task to program that type of insight — let alone the ability to verify the factual predicates of each situation — into a single computer, or even into a network system"(Pasquale, 2019, p. 26). Where systems are simple and variables are few, automation works. Where human complexity is introduced, as it invariably does in border control, welfare eligibility or policing, the automated system cannot account for what it cannot anticipate. The legal response to this condition, Pasquale argues, requires that the adage "a rule of law, not of men" be complemented by a new commitment: "a rule of persons, not machines" (Pasquale, 2019, p. 7). Without attributing algorithmic judgments to particular persons and holding them responsible for explaining those judgments, legal automation undermines basic principles of accountability (Pasquale, 2019, p. 7). This is a precise legal-theoretical statement of the condition Danaher identifies structurally. The rule of law presupposes identifiable persons who can be held responsible. Where algorithmic systems displace those persons, the accountability chain that democratic legitimacy requires is broken, not because the law has been violated, but because the architecture of decision-making has moved beyond the reach of the persons the law was designed to govern.
My analysis in this stage has drawn on five theoretical perspectives to frame the impact digital platforms have on democratic life. Habermas provides the normative baseline for what a non-distorted public sphere requires. Zuboff identifies the extractive commercial logic that structurally undermines it. Dean explains why that distortion remains invisible to those subject to it. Danaher frames the governance conditions that have produced an algocratic impasse. Pasquale identifies the legal- theoretical stakes of pursuing rule by machines rather than rule by persons. Together these accounts converge on a structural implication that none individually resolves. The systems identified do not just influence democratic life from outside. They exercise a form of power that is functionally analogous to the power of states (Gowder, 2023, pp. 8, 43).
In The Networked Leviathan (2023), Gowder argues that it "should be clear, perhaps even uncontestable, that platforms do carry out some tasks that are meaningfully similar, in both their social role and in the constraints they face, to tasks carried out by states"(Gowder, 2023, p. 43).6 To strengthen this, Srivastava identifies platforms as a form of private authority engaged in "authoritative decision-making that was previously the prerogative of sovereign states"(Srivastava, 2021, p. 990). This private authority is not simply a metaphorical device referencing state power; it is, in some contexts, replicating it, as platforms simultaneously challenge and reshape the traditional role of states while being deployed by existing geopolitical powers to extend and exercise their authority (Srivastava, 2021, p. 995). In the manner that states are formally equal but materially unequal, with some states capable of and willing to resist the constraints that smaller states cannot, platform power exists similarly. Larger platforms exercise their capacity to resist democratic accountability, challenging states’ authority to impose regulation. Srivastava identifies three distinct conditions through which this can be seen: interdependence, where states contract with and become dependent on Big Tech; circumvention, where platforms push against state overreach; and curtailment, where states attempt to regulate platforms (Srivastava, 2021, p. 995). These conditions can operate in parallel, often within the same state-platform relationship, creating a condition in which a state is simultaneously contractor and challenger.
The European Union provides a precise illustration. The EU has pursued Meta under the Digital Services Act and the General Data Protection Regulation, issuing significant financial penalties and demanding structural changes to algorithmic systems (European Commission, IP/24/2373, 30 April 2024). At the same time, European public institutions, businesses and citizens remain deeply dependent on Meta's infrastructure for communication, advertising and civic organisation. The same organisation is being regulated and relied upon simultaneously. Meta's response has been one of contestation—disputing findings, challenging jurisdiction and threatening service withdrawal. This is not the behaviour of a regulated private entity. It is the behaviour of an actor that understands itself as occupying a position from which democratic constraint can be negotiated rather than simply accepted. Meta has positioned itself as being strong enough, in a state-versus-state model, to challenge the EU. The accountability gap this produces is not a failure of regulatory design; such design is continuous in an attempt to match technological change. It is a structural condition generated by actors who occupy a governance position for which existing democratic frameworks were not designed.
Srivastava’s conditions operate at three simultaneous levels, each reinforcing the others. At the first level, platforms exercise quasi-sovereign governance over citizens, shaping the conditions of democratic agency through algorithmic systems that citizens cannot address. At the second level, platforms function as instruments of state governance, contracted by public authorities to carry out administrative functions whose logic remains proprietary and unaccountable to scrutiny. As Srivastava observes, one outcome of state-platform interdependence is that "large, global, privately-owned platforms become the regulatory agents of nation states"(Srivastava, 2021, p. 995). At the third level, platforms actively contest the authority of democratic institutions to govern them. They do this not as regulated entities disputing individual findings but as actors asserting the primacy of their own governance frameworks over democratic constraint. Together the three levels produce a condition that the preceding accounts identify in part, but do not fully resolve. The democratic problem is not just one of inadequate regulation; it is unaccountability operating concurrently at three levels: as quasi- sovereign governors of citizens, as instruments of state governance whose logic evades scrutiny, and as actors capable of contesting democratic authority itself.
Gowder's concern is that platforms lack the capacity to govern themselves responsibly. Platform self- control problems stem from “destructive methods of short-term profit seeking” and the dominance of engagement metrics over democratic considerations (Gowder, 2023, p. 4). The period between 2019 and 2021, during which platforms made their most assertive attempts at responsible self- governance—content moderation and election integrity measures, for example—represented an observable attempt at self-regulation. The systematic reversal of those commitments since 2022, across Meta, X and others, confirms precisely what Gowder voices concern about. Without external accountability mechanisms grounded in democratic legitimacy, platform governance defaults to commercially and politically expedient behaviour. His proposed solutions operate at the level of platform self-governance rather than democratic legitimacy itself. They address the symptom— inadequate capacity—without addressing the structural condition that produces it.
This claim finds independent agreement from a legal perspective. Klonick, examining the internal governance architectures of platforms, argues that they have become “new governors” developing bodies of rules, enforcement processes and adjudicative structures that mirror the architecture of public law (Klonick, 2018, p. 1663). Platforms do not simply influence the conditions of speech; they have built what resemble constitutional orders within which the conditions are defined, applied and contested. The content policies that determine what can be said, the manner in which they are applied and, most importantly, the appeals processes that adjudicate disputes are not market features. These are governance instruments, authored and applied opaquely, but operating over populations that, in some cases, are larger than the populations of states. Klonick's analysis is significant for the argument. She identifies the structural form of platform power rather than just scale. It is not that platforms are large, although scale provides the ability to close off justificatory routes en masse. It is that they have built the institutional architecture of governance without assuming the justificatory obligations that democratic governance requires. The end result is that “it leaves users essentially powerless” (Klonick, 2018, p. 1666).
Together, the approaches examined in this stage map the structural condition. I have established what is wrong at the structural level. The remaining question is normative: what obligations does that structural condition generate and what would discharging them require? A normative framework must be capable of specifying why algocratic rule, for example, generates democratic obligations and what those obligations would require. The question of what legitimacy demands of actors who exercise constitutive power over persons as political agents, as platforms do, and what follows when that demand goes unmet, requires a further move. That pivot is provided by Forst. Forst's concept of noumenal power shifts analysis to a complementary framework, underpinned by the discussion above, specifically the space of reasons within which political agency is formed and exercised.
Forst locates the real and general phenomenon of power not in physical force or institutional position but in what he calls the space of reasons, "the realm of justifications"(Forst, 2017, p. 38). To be a subject of power, on this account, is to be moved by reasons that others have provided, reasons that motivate thought and action in ways intended by the reason-giver. Power, in this sense, is not simply exercised by and over free agents; it is the structuring of the conditions within which those agents form their reasons in the first place. The precise formulation Forst offers is "to have and to exercise power means to be able — in different degrees — to influence, use, determine, occupy, or even seal off the space of reasons for others"(Forst, 2017, p. 42). This is a definition of power that operates prior to the observable exercise of force or institutional authority. It locates power at the architectural level, the level at which the conditions of reasoning itself are formed. The implications for understanding how power works are significant. As Forst argues in The Noumenal Republic (2024), "we fail to understand how power is exercised … if we do not understand the justifications that guide people's actions"(Forst, 2024, pp. 122-123). This is a direct claim about where power actually resides—in the structuring of justificatory conditions rather than in the observable outputs those conditions produce. It exists prior to whatever generates those outputs.
The structural condition identified throughout Stage 2 operates at precisely this level. What Forst provides is the normative explanation for why that condition is democratically illegitimate: not merely harmful, though it may be, nor simply inadequately regulated, as the current position in 2026 reflects, but a violation of the conditions that democratic legitimacy itself requires. This normative explanation rests on Forst's principle of justification. Those subjected to a normative order, Forst argues, “should be its co-authors as equal participants and normative authorities in adequate justificatory practices that critically reflect on and constitute that order"(Forst, 2017, p. 42). This is the first demand of justice— to have standing as an equal normative authority within the order that governs you. The democratic expression of this principle is stated directly in his 2024 work: "democratic rule exists where those who are subjected to a normative order are at the same time the normative authorities who co- determine this order through corresponding justification procedures"(Post, 2024, p. 187). Democratic legitimacy is not captured by elections, representation or the formal procedures of horizontal and vertical accountability. It requires that those subject to the conditions that structure their political agency are able to contest and co-determine those conditions. Where that requirement cannot be met—where the space of reasons is occupied, shaped or sealed off by systems that those subject to them cannot inspect, contest or influence—democratic legitimacy is not diminished or weakened. It is structurally precluded.
This normative framework identifies a condition that the preceding accounts have mapped but not fully resolved. Algorithmic systems occupy and shape the space of reasons within which citizens must be allowed to become political agents. The systems do this without being subject to the justificatory obligations that democratic legitimacy requires of any agent that exercises this kind of power. The obligation is not discharged by horizontal accountability mechanisms alone, by ombudsmen or courts. Such mechanisms operate at one remove from the individuals whose space of reasons is being structured. What Forst's principle of justification requires is that individuals themselves have meaningful standing as normative authorities in relation to the systems that govern their political agency, that the architectural conditions of democratic life are themselves open to justificatory challenge by those subject to them. I propose a ninth dimension, adding to Diamond and Morlino's framework and designed to capture the gap created by algorithmic systems. It is not concerned with what individuals and groups do within democratic life; that is the domain of Diamond and Morlino's existing eight dimensions. It is concerned with the conditions under which democratic life becomes possible in the first place. Where algorithmic systems occupy the space of reasons, shaping their world view, without being subject to the justificatory obligations that Forst's principle requires, those conditions are structurally compromised. The existing dimensions remain capable of measuring democratic outputs. What they cannot measure, and what the ninth dimension addresses, is whether the architectural conditions that make those outputs democratically meaningful are themselves subject to the justificatory challenge that democratic legitimacy demands. That is the gap I identify. That is the gap the ninth dimension is designed to close.
Stage 2 established that algorithmic systems exercise a form of noumenal power that shapes democratic activity before actions occur, and that the obligation of justification this generates cannot be discharged through horizontal accountability mechanisms alone. This stage applies the argument against three of Diamond and Morlino’s dimensions: participation, responsiveness and the rule of law. They represent, respectively, the inputs, outputs and legal foundations of democracy. What the analysis reveals is not three separate problems but a single limitation expressing itself in three different registers. The dimensions do not fall short. They continue to function as normative and empirical measures of democratic quality, as analytically strong as they were designed to be in 2004. What they cannot do is reach the architectural level at which the conditions of democratic agency are formed before observable activity begins. The gap lies not within any single dimension but in what all of them presuppose: a background environment in which the conditions for democratic agency can be taken as given. It is here that the framework’s reach ends and the case for an additional dimension begins.
Diamond and Morlino's participation dimension measures the formal capacity of individuals to engage in democratic life: voting, organising, petitioning and making demands of power.7 A good democracy, they argue, must ensure that citizens are in fact able to make use of these formal rights to influence the decision-making process (Diamond and Morlino, 2004, p. 23). These are observable activities, measurable through voting data returns and more independent indices such as V-Dem. The dimension works on the basis that the conditions under which those formal rights are exercised remain broadly intact. If citizens are prevented from voting through identification requirements, the dimension can register and account for that constraint. Where there is a one-party state with nominal elections, this can be factored into a judgement on how participatory the process genuinely is. The architectural argument puts the dimension under pressure in scenarios where there is apparent freedom to vote, a range of parties and no outward coercion. On its face, this looks like a neutral environment. In 2026, however, there is an unseen determinant impacting participation prior to citizens making informed choices. This is the condition Habermas identifies when he argues that algorithm-controlled platforms are 'far from neutral' in their operation, distorting the communicative conditions on which political agency depends (Habermas, 2022, p. 163). What platforms appear to offer as open participation diverges, internally, from what they deliver. The divergence operates before any act of participation takes place.
The divergence Habermas identifies highlights an aspect of participation the dimension cannot resolve on its own. An examination of a counterfactual scenario highlights this limitation: if the shaping of participatory agency were just an additional form of observable constraint, not present in 2004, such as voter identification requirements or one-party competition, it could in principle be factored into an assessment of participatory quality. It would form one of the variables applicable to measuring the quality of participation specifically and democracy more generally. There would be no need for a ninth dimension. My argument is that the architectural condition is categorically different from those the dimension was designed to capture. It does not constrain participation from outside. It fabricates the conditions of participation from within, before any observable activity takes place. That is a different kind of problem, requiring a different kind of analytical approach. High voter turnout, for example, registers as positive engagement within Diamond and Morlino's framework. Where that turnout has been shaped by algorithmically curated information, however, the distinction between acting and being moved to act becomes structurally obscured. While the observable trajectory remains identical to an external observer, the underlying political agency is fundamentally distinct. The direction is the same, the agency is not. In a simplified model, the dimension records the output as a percentage of outcomes. It cannot, on its own, assess the conditions that produced the outcomes. As Alnemr (2025, pp. 2-3) demonstrates, algorithmic systems promote political polarisation and compromise the conditions of political agency before citizens enter the public sphere at all. Dean's concept of the ‘fantasy of participation’ explains why this is so difficult to detect. Citizens believe they are genuinely active. The experience of using the digital environment is indistinguishable, from within, from meaningful democratic participation. What the dimension cannot assess is the degree to which the architecture has pre-structured what citizens encounter and how their demands are formed. Dean describes this as a structural condition of depoliticisation in which the performance of participation substitutes for its substance (Dean, 2005, pp. 60-61). Participation may be formally robust while its conditions are being structurally hollowed out and the dimension has no means for detecting the difference.
My argument does not depend on endorsing any particular theory of what participation should look like. My own position sits toward the thin end, where a minimum of features would be in place. That leaves open further discussion as to whether participation is of a participatory, deliberative or liberal democratic nature. The architectural problem here cuts across theories of democratic theory. This matters for the strength of my claim. Deliberative democracy requires the conditions for reasoned public exchange. As established, those conditions are structurally compromised at the architectural level rather than simply degraded at the output stage. Representative democracy requires that citizens form preferences independently and hold governments to account through the mechanisms of election and representation. Where algorithmic systems pre-structure those preferences before any institutional engagement occurs, the representational chain is mediated at its source: the government that is elected, and the policies it pursues, reflect an architecturally shaped input rather than an independently constructed one. Participatory democracy, even in its thinner forms, requires only that citizens can meaningfully engage in political life through the available channels. Where the conditions of engagement are structurally invisible, participation is hollowed out regardless of how many channels formally exist. Any empirical data the dimension generates about participation loses democratic meaning in this condition, rendering discussion about which type of democracy is in operation premature. The architectural problem is prior to any dispute about which theory of democracy is correct. It operates at the level of the conditions that concepts of democracy presuppose without examining.
Applying a deliberative framework, Castelló et al. (2026) attempt to address this problem. They reveal where that approach reaches its limit, arguing that social media platforms have evolved into commodified spaces whose business models incentivise hate speech, misinformation and political fragmentation. Their response is to enhance platforms' deliberative capacity across four dimensions: transparency and accountability, openness and inclusiveness, conduciveness to argumentation and consequentiality (Castelló et al., 2026, pp. 12-15). Their four ‘capacities’ describe, in deliberative terms, what a democratically legitimate platform environment must provide. Each, however, presupposes what the architectural argument identifies as structurally absent. Conduciveness to argumentation requires that users engage in reasoned exchange rather than merely circulate content, their “essential dialectics” (Castelló et al., 2026, p. 14). However, the architecture determines what counts as engagement before any dialectic begins. Consequentiality requires that deliberation connects to collective decision-making in ways that citizens can recognise and contest. Problematically, where the architecture shapes both what citizens understand themselves to want (the 2026 problem I have identified) and what they understand themselves to be entitled to demand, the chain between deliberation and consequence is broken. Their institutional recommendation, that platforms be reconceptualised through a social contract between platform firms and stakeholders (Castelló et al., 2026, p. 16), underestimates the structural asymmetry at stake. As established in Stage 2, platforms are not parties to a contract in any conventional sense. They, increasingly, exercise quasi- sovereign power while simultaneously contesting the authority of democratic institutions to govern them. A social contract presupposes roughly equal parties capable of negotiating terms. That presupposition does not currently hold, and until it does, the deliberative capacities Castelló et al. identify remain aspirational rather than operational. What their framework cannot provide is a standard for the architectural conditions that must be in place before any of those capacities can function. That is the gap the ninth dimension is designed to address.
The shaping of participatory conditions is not confined to social media, but it is most often the aspect which is highlighted. It operates across the full range of digital systems through which political life is now mediated. Google's search ranking determines how political information is ranked and delivered. YouTube's recommendation system shapes what political content is consumed and in what sequence. News aggregators such as Apple News and Google News curate what counts as relevant journalism, algorithmically determining which stories reach which citizens and in what order. Each produces a personalised information environment that fragments the shared epistemic foundation on which democratic participation depends. What one citizen receives as politically true or legitimate may bear little resemblance to what another encounters, not because of differences in their reasoning, but because of differences in what their algorithmic environment has made available to them.8 Helberger's analysis of news recommender systems identifies the democratic stakes of this condition directly. She shows that recommender algorithms are built around assumptions about what kind of democratic citizen they serve and that systems developed with commercial objectives rather than democratic values in mind systematically fail to provide the conditions that participatory or deliberative democracy requires (Helberger, 2019, pp. 997, 1009). More fundamentally for democracy, the architecture of recommendation encodes a model of citizenship before any content reaches those it shapes. Pariser's filter bubble (2011, pp. 9-10) and Sunstein's echo chamber (2017, pp. 9-11) diagnose this condition at the level of content. While the growth of harmful content remains a pressing empirical concern, my architectural argument advanced here operates at a conceptually prior level. By the time the content is delivered it is too late. The architecture shapes all individuals who do the seeing before any content arrives.
The upward direction is not, however, the only way in which the participation dimension is placed under architectural pressure. Participation operates in a second direction, beyond the public sphere. Where the first direction concerns the upward flow of citizen agency, the second concerns the exercise of state power through algorithmic systems that participation itself does not reach. Vertical accountability requires that the state's exercise of power over individuals remains visible, attributable and challengeable. Where it does not, citizens cannot contest decisions that affect them, representatives cannot examine the actions of government and the relationship between state and citizen is hollowed out from within. As Stage 1 of this investigation established, both the Palantir contracts and the DWP's Universal Credit Advances Model illustrate this condition. In each case, the systems that govern citizens' relationship with the state operate under Freedom of Information exemptions and proprietary protections that place their logic beyond public reach. The Public Law Project's TAG register identifies 55 active algorithmic systems across UK public administration, the majority operating with limited transparency (Public Law Project, 2024). This condition is not exceptional. It is the way in which public administration now routinely operates. The consequence for participation is structural, removing an avenue through which participation and accountability are traditionally linked. A citizen may petition against the deployment of machine learning in welfare decisions, as over 300,000 did in relation to Palantir contracts in April 2026, while the systems that govern their relationship with the state continue to operate unchanged. The formal channels of participation function. The substantive connection to the addressee that matters is removed. Petitioning against a system whose logic cannot be inspected is participation directed at a locked digital cage.9 The participation dimension has no mechanism for detecting this condition. It registers the petition. It cannot register the architectural opacity against which the petition is directed. In Forst's terms, the individual retains the phenomenal status of a participant while being noumenally reduced to a data point: present in the count, absent from the conversation (Forst, 2017, p. 42).
Responsiveness asks whether governments respond to the wants of citizens. Diamond and Morlino, using Powell, identify “three links in the chain of democratic responsiveness”: preferences are formed and structured by competing political parties, aggregated through electoral processes into a government of policy makers and translated by elected officials into policy outcomes (Diamond and Morlino, 2004, pp. 27-28). As a measure of democratic quality, it works well across a range of democratic styles. For responsiveness to function as a democratic measure, two conditions must hold. First, the preferences and demands to which governments respond must be formed independently, by citizens exercising genuine deliberative agency. Second, those preferences must be directed toward an identifiable addressee, a government or institutional arm, capable of receiving and acting on them. Where algorithmic systems intervene, shaping both what individuals take themselves to want and curating what they understand themselves to be entitled to demand, both conditions are compromised. Powell's first link in the chain, preference formation, is compromised before the chain begins. The commercial logic of engagement maximisation results in the information environment, through which preferences are formed, being optimised not for deliberative quality but for retention and reaction. An environment optimised for retention and reaction rather than considered exchange is not one in which independent preferences can reliably form. These are not typical features of a responsive process. As Castelló et al. (2025, pp. 1-2) demonstrate, platform architectures systematically undermine information quality, creating structural power asymmetries that shape what citizens understand themselves to want before any formal political demand is articulated. The architecture registers both as engagement data, weighting content by its capacity to retain attention rather than its democratic quality. A government responding to the preferences this environment produces can score highly on Diamond and Morlino's responsiveness dimension while responding to something that is, at least in part, an architecturally pre-formed output rather than an independently constituted political demand. Responding to algorithmically amplified pressure instead of considered preference removes the space for longer term deliberation, further eroding the conditions under which genuine democratic responsiveness can operate.
Where algorithms systematically prioritise content that provokes strong emotional responses, the democratic consequence is that responsiveness risks becoming structurally biased toward populist outputs. This is not a contingent risk erasable through governance. It is a structural tendency built into the logic of the platforms through which political preferences are increasingly formed. Maly identifies this condition as 'algorithmic populism': the process through which algorithms function not as neutral distributors but as active co-constructors of political identity, producing 'the people' as an engagement-optimised construct rather than an independently formed democratic constituency (Maly, 2018, pp. 8, 13-14). The architecture rewards engagement over deliberation, with engagement metrics substituting for genuine public support (Maly, 2018, p. 9; Rădulescu, 2025, p. 9). Empirical evidence confirms the directional character of this tendency and its political direction. Huszár et al. (2022, p. 2), in a large-scale randomised experiment involving two million daily active Twitter accounts across seven countries, found a consistent trend: in six out of seven countries studied, the mainstream political right enjoys higher algorithmic amplification than the mainstream political left. The study is notable because it was conducted in part by Twitter's own Machine Learning Ethics, Transparency and Accountability Team, making it a case of a platform's own researchers confirming the amplification bias built into its architecture (Huszár et al., 2022, p. 1). More recently, Ye, Luceri and Ferrara's (2025, pp. 1, 5) audit of X's recommendation system during the 2024 US Presidential Election found that neutral new accounts show a default right-leaning bias in content exposure and that both left and right-leaning accounts encounter amplified exposure to content aligned with their own political stance and reduced exposure to opposing viewpoints.10 A government responsive to this environment is not responding to its citizens. It is responding to an algorithmically curated and directionally biased approximation of them. Where that approximation systematically amplifies populist and emotionally charged content, responsiveness becomes a transmission mechanism for architecturally produced political pressure rather than a measure of democratic quality.
UK immigration policy illustrates the dangers of influencing responsiveness. Increased government responsiveness to anti-immigration sentiment registers positively within Diamond and Morlino's framework. The current Labour government has it as a major agenda point, as did previous governments. What the dimension cannot capture is the extent to which that sentiment has been amplified and channelled by recommender systems optimised for engagement rather than reflective judgement. The 2024 summer unrest provides a concrete illustration of the mechanism. Following the Southport murders of July 2024, the recommendation algorithms of social media companies amplified misleading and hateful messaging, driving protests that turned violent and targeted Muslim and migrant communities. The House of Commons Science, Innovation and Technology Committee found that social media business models incentivised the spread of misinformation and that algorithmic recommendations played a role in driving divisive narratives during the crisis period (House of Commons Science, Innovation and Technology Committee, 2025). Amnesty International's subsequent analysis of X's open-source recommender algorithm identified systemic design choices that favour contentious engagement over safety, concluding that the platform's algorithmic design and policy choices contributed to heightened risks during the wave of anti-Muslim and anti-migrant violence (Amnesty International, 2025). The amplification was not incidental. It was architectural. The longer-term shaping of anti-immigration sentiment through platform dynamics is documented in the trajectory of Reform UK and Restore, whose rise has been partly attributed to algorithmic amplification on Meta and other platforms, functioning as accelerants for populist sentiment that pre- exists the architecture but is normalised and scaled through it. The government's subsequent policy responses, including accelerated deportation measures and a hardened public tone on asylum, represent responsiveness to a political environment that has itself been architecturally shaped. The policy response may be entirely genuine. The preferences it responds to are not entirely independent. Responsiveness, in this condition, risks becoming a measure of governmental sensitivity to algorithmically curated public pressure rather than to the considered preferences of democratic citizens.
If, as Diamond and Morlino argue, the rule of law is "the base upon which every other dimension of democratic quality rests" (Diamond and Morlino, 2004, p. 21), limitations identified here carry greater consequences. A structural inadequacy in the rule of law dimension does not affect only that dimension. It undermines the normative foundation on which the remaining seven rely. My argument in this section is that algorithmic systems introduce amplified inadequacy, not by violating the rule of law on its own terms, where existing statutes and guardrails provide partial mitigation, but by operating at a level that the law has not yet been designed to reach. In 2025, 4.2 million people's faces were scanned by UK police using live facial recognition technology, more than in any other European capital or Western democracy (Big Brother Watch Team, 2026). The de jure rule of law remained intact. No explicit statutory provision was violated. The de facto reality, however, preceded the legal framework: individuals were subject to forms of algorithmic identification they were unable to inspect, challenge or refuse prior to their operation. The High Court challenge brought by Shaun Thompson and Silkie Carlo in January 2026 confirms the point. Legal challenge was available, but only after the architectural condition had already taken effect. The issue is not illegality in the conventional sense. It is a structural inadequacy: the framework of law exists, but it does not yet reach the space in which power is being exercised. This structural inadequacy was confirmed, paradoxically, by the court's own verdict.
The High Court dismissed the claim, ruling the Metropolitan Police's updated deployment policy to be in accordance with the law and prescribed by law under Articles 8, 10 and 11 of the European Convention on Human Rights (R (Thompson and Carlo) v Commissioner of Police of the Metropolis [2026] EWHC 915 (Admin), para 229). The court found that internal safeguards, including mathematically mapped crime hotspots and mandated administrative authorisation processes, satisfied the qualitative requirements of foreseeability under the Convention (paras 95-98, 127, 229). Crucially, however, the court reaffirmed that its supervisory jurisdiction in judicial review is strictly restricted to checking whether a public body acts within its legal powers, not to examining the underlying operational merits of the technology itself (para 5). By ruling that a rigorous internal procedural trail is sufficient to satisfy human rights standards (paras 217-227), the court reduces the rule of law to procedural compliance. The inner mathematical logic of the algorithm, the training datasets and the corporate operational experience driving deployment selections remain a closed domain insulated from public scrutiny or judicial oversight (paras 194-199, 184). The High Court's verdict structurally instantiates the very noumenal power imbalance established in Stage 2, transforming the rule of law into an administrative shield for automated state action rather than a mechanism for democratic contestation. Further, the Metropolitan Police settled a damages claim brought by Shaun Thompson arising from his wrongful identification, even as the court upheld the policy's legality (para 27; Big Brother Watch Team, 2026). The architecture passed the legal test while producing a compensable wrong. De jure rule of law was satisfied. The de facto harm had already occurred.11 When evaluated through Diamond and Morlino's framework, this verdict exposes the vulnerability of the rule of law when confronted with automated power. For Diamond and Morlino, a robust rule of law requires transparency, equality and predictability so that citizens can understand and legally check how state power is being exercised. The court's reasoning satisfies the formal conditions of that standard while leaving its substantive requirements unmet. The claimants have announced their intention to appeal, confirming that the architectural problem identified here remains live and unresolved.
Pasquale's analysis of legal automation sharpens the structural inadequacy the Thompson and Carlo verdict exposes. He argues that the removal of identifiable persons from decision-making processes breaks the accountability chain that democratic legitimacy requires (Pasquale, 2019, pp. 5-6). The rule of law, on this account, is not simply a set of substantive constraints on state power. It is a relational structure in which power is exercised by persons who can be identified, questioned and held responsible. Where algorithmic systems replace that relational structure with automated processes, the law does not fail on its own terms. It is circumvented at a level outside the reach of its own mechanisms. The Metropolitan Police's live facial recognition deployment illustrates the condition. No identified officer decided to scan Shaun Thompson's face. A system did. The legal challenge that followed had to be directed at the policy authorising the system rather than at any person who had exercised judgment about him. The accountability chain was not broken. It never existed. The architecture replaced the person before the law had occasion to reach either.
Erickson and Gregory's analysis of algorithmic specification identifies how deep the problem is. Where Pasquale identifies the displacement of persons, Erickson and Gregory identify the displacement of legal concepts themselves. When legal ideas are translated into computational form, the translation does not simply automate the application of existing meaning. It transforms the meaning itself. Terms that were designed to remain interpretively open, accommodating moral development and contextual judgment, are converted into fixed parameters, closing down the interpretive space that legal language was designed to preserve and producing a conceptual rigidity that the law's normative function cannot survive (Erickson and Gregory, 2025, p. 5). The authority to determine what legal concepts mean passes from courts and legislatures to the engineers and training datasets that define the system's inputs, an unauthorised reassignment of interpretive authority from legal institutions to technical systems (Erickson and Gregory, 2025, p. 8). The problem does not stop there. As algorithmic systems are deployed repeatedly, their outputs begin to shape institutional expectations and narrow what counts as a legitimate interpretation, feeding back into the legal system and influencing future practice (Erickson and Gregory, 2025, p. 5). The algorithm becomes, as Erickson and Gregory put it, an "unacknowledged legislator, shaping legal meaning under the guise of technical necessity" (Erickson and Gregory, 2025, p. 4). At that point the rule of law is not merely formally present while substantively hollow. It is being gradually displaced by a parallel architecture of meaning, one that runs alongside the human system while quietly supplanting it. In the live facial recognition case study, the concept of lawful identification is no longer interpreted by an officer exercising judgment in relation to a person. It is computed by a system whose classificatory logic is neither publicly articulated nor judicially reviewable and whose repeated deployment is quietly generating the normative categories by which future deployments will be assessed. The rule of law is formally present. Its interpretive substance has been evacuated.
The three dimensions examined in Stage 3 share a single structural problem. Each continues to function as a normative and empirical measure of democratic quality. What none of them does is reach the level at which the conditions of democratic agency are formed before observable activity begins. The participation dimension cannot detect the architectural fabrication of participatory conditions. Responsiveness cannot distinguish independently formed preferences from algorithmically curated approximations of them. Rule of law cannot reach the level at which legal concepts are transformed by their encoding into systems whose inner logic remains closed to judicial and democratic scrutiny. These are not failures of the dimensions. They are the boundary of the framework's reach. I argue that boundary marks the space a ninth dimension must occupy, in 2026 and beyond: one capable of measuring the architectural conditions that determine the quality of democratic agency before any of the existing eight dimensions begin their work.
The proposed ninth dimension is Algorithmic Accountability. It is concerned with the extent to which the systems that position individuals as political agents are themselves answerable to those they govern. Unlike the original eight dimensions, Algorithmic Accountability does not measure what individuals and groups do within democratic life: that remains the scope of the 2004 framework's normative architecture. It addresses the conditions under which democratic life becomes possible in the first place. The distinction between these two levels is not simply analytical. Diamond and Morlino's eight dimensions, taken together, function as a normative architecture designed to keep the space of reasons open. Fair elections, rule of law, participation, responsiveness, horizontal and vertical accountability: each operates as a structural guarantee against the kind of pre-agency closure that would undermine democratic legitimacy. In 2004, that architecture was sufficient because the means of exercising constitutive power over the space of reasons at scale did not exist outside the reach of those dimensions. Any actor capable of closing off the conditions of political agency would have had to operate through the phenomenal surface of democratic life to do so, and, in doing so, would have become visible to the framework's measures. The framework does not guarantee against such closure, as the cases examined across this investigation confirm, but it provides an identifiable and, potentially, judicially accountable focus for democratic challenge.
Consider a thought experiment. In 2004, as in 2026, the conditions for democratic collapse existed. A political actor, 'General Jones', manages to secure elected office. Once in power, he dismantles horizontal accountability bodies, packs the senior judiciary with compliant appointments, closes media outlets hostile to his administration and repeals legislative protections for minorities. Every action General Jones takes is normatively legible within Diamond and Morlino's framework. The disbanding of accountability institutions registers against the accountability dimension. Judicial capture registers against rule of law. Media closure registers against freedom. The repeal of minority protections registers against equality. The framework catches all of it, because General Jones, however authoritarian his intentions, must operate through the phenomenal surface of democratic life. His actions are visible, attributable and leave measurable traces across the eight dimensions. He can be identified, named and made the potentially judicially accountable focus of democratic challenge. All of this would feature in registers such as V-Dem, which tracks precisely these dimensions of democratic quality across the full range of observable institutional behaviour (Nord et al., 2026). What General Jones is doing, in Levitsky and Ziblatt's terms, is "capturing the referees": targeting the judicial system, law enforcement bodies and regulatory agencies whose function is to serve as neutral arbiters of the democratic process (Levitsky and Ziblatt, 2018, p. 79). Alberto Fujimori's presidency in Peru illustrates the pattern with particular clarity. Elected democratically in 1990, Fujimori progressively dismantled the institutional architecture around him, dissolving Congress and the constitution in 1992, packing the courts with loyalists and using intelligence services to surveil and blackmail political opponents. Each step was formally legal or presented as an emergency measure. Each left a measurable trace across the dimensions of democratic quality. Each had an identifiable author. The referees were captured and the capture was measurable (Levitsky and Ziblatt, 2018, ch. 4).
Donald Trump's second term provides a contemporary illustration of the same dynamic operating within a mature democracy, subject to the 2004 dimensions but also impacted by algorithmic distortion. Since his inauguration, the administration has systematically dismantled horizontal accountability structures across multiple dimensions simultaneously. The Department of Government Efficiency (DOGE) dismantled federal oversight bodies and purged inspectors general across nineteen positions (Nord et al., 2026, p. 35). Executive orders targeted specific law firms whose lawyers had pursued cases against the administration, directly threatening the independence of the legal profession (Trump, 2025). The administration repeatedly defied court orders and publicly intimidated judges. Freedom of expression in the United States, on V-Dem's measure, fell to its lowest level since the end of the Second World War. The V-Dem Democracy Report 2026 characterises Trump's second term as "a rapid and aggressive concentration of power in the presidency"(Nord et al., 2026, p. 1) that has undercut institutionalised checks and balances, politicised the civil service and intimidated the judiciary, recording the largest single-year drop in American democratic standing in the entire history of their dataset (Nord et al., 2026, p. 20). All of these actions are normatively legible within Diamond and Morlino's framework. The actors are identifiable, mechanisms are constitutionally visible and the democratic consequences are measurable. General Jones, in his contemporary form, is captured by the eight dimensions.
Now run the same thought experiment with algorithmic systems doing equivalent work. The space of reasons is constituted before political agency forms. Preferences are architecturally designed before citizens enter any accountability process. The judiciary cannot reach the level at which power is being exercised, as Thompson and Carlo confirms (R (Thompson and Carlo) v Commissioner of Police of the Metropolis [2026] EWHC 915 (Admin), para 229). Minorities are disproportionately flagged by algorithms and media plurality rules do not apply. Here, there is no General Jones undermining democracy. There is, in its place, a lack of agency. Nevertheless, there is a documented pattern of conditioning that shapes the political environment within which General Jones becomes possible. Huszár et al.'s finding that the mainstream political right enjoys higher algorithmic amplification than the mainstream political left across six of seven countries studied, confirmed by Twitter's own research team, identifies a structural tendency built into platform architecture before any individual actor exploits it (Huszár et al., 2022, p. 1). Ye, Luceri and Ferrara's audit of X's recommendation system during the 2024 US Presidential Election found that neutral new accounts show a default right- leaning bias in content exposure, with both left and right-leaning accounts encountering amplified exposure to content aligned with their existing political stance (Ye, Luceri and Ferrara, 2025, pp. 1, 5). The architecture did not create Trump. It constituted and sustained the conditions within which the preferences, beliefs and political demands that brought him to power were formed, normalised and insulated from counter-deliberation before any institutional accountability mechanism had occasion to engage. Once in power and backed by a participatory mandate, General Jones begins the deconstruction of democracy. What distinguishes the contemporary version is not the intent but the architecture: algorithmic mechanisms that amplify, accelerate and conceal what previous autocrats could only attempt through visible institutional force. The eight dimensions are still measuring the consequences. However, they have no mechanism for assessing the architectural conditions that shaped and continue to shape the preferences and beliefs that brought him there, or that continue to insulate his political support from the reflective challenge democratic deliberation requires. The framework has not failed here. It is doing exactly what it was designed to do. What it was not designed to do, because in 2004 no agent possessed the capacity, was govern the architectural conditions that precede its own operation. The threat of pre-agency closure was latent in 2004. The eight dimensions were normatively sufficient for that environment because the technical capacity to exercise constitutive power over the space of reasons at democratic scale did not yet exist outside their reach. Any actor seeking to achieve such closure would have had to pass through the phenomenal surface of democratic life to do so, becoming visible to the framework's measures in the process. What has changed is not the normative logic of the framework. It is the emergence of systems capable of exercising constitutive power over the space of reasons at a level the eight dimensions were never designed to reach. The ninth dimension addresses that gap. It provides the link between the original normative requirements of the 2004 framework and the architectural conditions of 2026 political life: the level of analysis at which the preconditions of the other eight dimensions can themselves be examined and, where necessary, contested.
Forst's principle of justification requires that any agent exercising constitutive power over persons as justificatory beings must answer to those it governs (Forst, 2024, p. 2). Diamond and Morlino do not describe their framework in noumenal terms, but their account of democratic quality already depends on conditions that are noumenal in structure: conditions that shape the space within which citizens can appear as equal political agents and justificatory beings. Their dimensions are not merely a catalogue of visible institutional outputs; they presuppose a deeper normative order in which equality, agency and contestability must already be secured if democracy is to function at all. The ninth dimension makes that implicit level explicit. It names the architectural conditions under which justificatory claims can be formed, directed and received, and it does so because the existing dimensions only make sense if those background conditions are in place.
My noumenal claim requires demonstration through application. Consider three of Diamond and Morlino's dimensions directly. The equality dimension measures whether citizens are treated equally under the law and whether equal political rights are formally protected. For equality to function as a democratic measure rather than a simply procedural one, it must presuppose that those who are formally equal are also equal as normative authorities. Their claims carry the same justificatory standing and the political order is answerable to them on equal terms. This is not a contingent addition to the equality dimension. It is what makes equality democratically significant rather than administratively convenient. The participation dimension measures the capacity of citizens to engage in democratic life: to vote, organise, petition and make demands of power. For participation to function as a measure of democratic quality rather than mere activity, its anchor point is that citizens are genuine agents whose engagement reflects independently formed judgements rather than structurally pre-determined outputs. Without agentive capacity, participation is performance rather than practice. In this case, the dimension has nothing meaningful to measure. The accountability dimension, which Diamond and Morlino identify as requiring citizens to hold governments responsible for their decisions, presupposes that citizens can form coherent demands, direct them toward an identifiable addressee and receive a substantive response. That is a reason-giving relationship in the precise sense Forst identifies: a structured exchange in which one party exercises power and is obligated to justify that exercise to those subject to it. None of these presuppositions are Forstian by name. They are Forstian in structure. Each identifies a condition of justificatory agency— equal standing, agentive capacity and reason-giving addressability—without which the dimension in question ceases to measure democratic quality and measures only observable behaviour. Forst does not supply these presuppositions to Diamond and Morlino from outside. He names what was already structurally required for their framework to do the normative work it was designed to do.
Specifying that obligation in practical terms requires a choice between what Williams distinguishes as thick and thin normative standards (Williams, 2006, pp. 140-142). A fully thick standard would require that every architectural decision be open to justificatory challenge and co-determined by those subject to it. Such a standard is neither operationally achievable nor analytically useful as a dimensional measure. It would be impossible to determine when it had been met, and it risks collapsing into an indeterminate ideal that generates no measurable criteria for comparative assessment. The ninth dimension operates instead with a deliberately thin normative standard. This would provide a transferable framework suited to governance and legal structures globally rather than specific to one region. Thin, in this context, does not mean weak. A thin democracy is one that retains the minimum features and structures necessary before it ceases to be democratic at all. The equality and agency presuppositions embedded in Diamond and Morlino's dimensions provide the guidelines for what a thin standard requires. Equality, in its thin sense, does not mean that all citizens receive identical treatment or identical outcomes. It means that all citizens possess the basic status of persons whose justificatory claims must be capable of being made and received. This is a precondition that must be in place before and regardless of whether any claims are made. Agency, in its thin sense, does not mean that all citizens exercise identical degrees of political influence; clearly they do not. It means that all citizens possess the structural conditions necessary to form, direct and contest political claims as genuine normative authorities. This too is a requirement regardless of whether any or all citizens choose to exercise their agency. Where citizens make competing claims, democracy's function is to adjudicate between them. Regardless of outcomes, there is an assumption that the justification for successful and unsuccessful claims is traceable and examinable under a democratic lens. The thin standard therefore specifies only what must be in place before either of these conditions can function at all. It leaves open the further and contested question of what full democratic equality or full democratic agency would look like. That question is not resolved here. It is, however, made askable for the first time by the ninth dimension. For the purposes of this dimension, the thin standard is defined by three requirements. Each is derived from Forst's principle of justification and from the structural conditions identified across Stages 1 to 3. Together they provide the basis for empirical operationalisation without prescribing a particular vision of platform governance.
The first requirement is architectural transparency. For justificatory agency to be possible at even its thinnest level, individuals—citizens, civil servants, judiciary—must be able to identify the governing logic of the systems that shape their political environment with sufficient clarity to contest it. This does not require full algorithmic disclosure. At the thick end, that would lead into technical accounts beyond the reach of most citizens and most governance frameworks. The standard I apply is closer to the democratic principle already embedded in representative practice: a citizen engaging their MP to table a motion does not need the MP to explain parliamentary procedure in full. What matters is that the channel exists, is accessible and produces a genuine response. The same minimum applies here. Brauneis and Goodman's study of algorithmic transparency in the US public sector provides a wide- ranging account of what this requires and why it so frequently fails. Filing 42 open records requests across 23 US states for information on six predictive algorithmic systems, they found that in almost every case meaningful algorithmic transparency was not provided (Brauneis and Goodman, 2018, pp. 103, 151-152). Their concept of meaningful transparency is precise and directly applicable here: not source code disclosure, which they argue is rarely necessary, but knowledge sufficient to evaluate the utility and fairness of the system, and to understand what policy judgments the algorithm embodies (Brauneis and Goodman, 2018, p. 132). The failure to provide even this minimum is problematic in the public sector. Democratic governments bear duties of accountability and, because the risk of algorithmic opacity is not merely epistemic but political, it creates the conditions for what they term "corporate capture of public power", where private vendors make design and policy choices out of the sight of client agencies, the public, or both (Brauneis and Goodman, 2018, p. 109). Engelmann reinforces this from a constitutional perspective, arguing that algorithmic transparency is not simply a governance requirement but a fundamental right in the democratic rule of law, derived from the constitutional principles of due process and accountability: "the possibility of overseeing activities conducted by the State, an intrinsic element of democracy itself"(Engelmann, 2023, p. 173). Toledo's review of 78 publications reinforces that the structural problem Brauneis and Goodman identified in 2018 remains: algorithmic systems continue to operate as opaque computational black boxes whose internal decision logic is difficult to interpret, creating persistent challenges for transparency and oversight because citizens, policymakers and auditors cannot determine how automated systems reach specific decisions (Toledo, 2026, p. 3). The formal equality of citizens as political agents is unchanged at the phenomenal surface while being hollowed out at the architectural level. The competing claims that democracy is designed to adjudicate cannot reach the justificatory level at which they are being pre-structured. Without transparency of sufficient depth, the first condition of the thin standard cannot be met.
The second requirement is justificatory addressability. When a system shapes the conditions of political agency, there must be somewhere to direct a challenge. Not a complaints form or a regulatory body acting on a citizen's behalf, but a genuine relationship of accountability between the system and those it governs: an identifiable person or body in authority capable of receiving and responding to justificatory claims. Hillo, Vento and Erkkilä's experimental evidence, drawn from randomised survey experiments involving 842 senior public administrators and 3,245 citizens in Finland, provides empirical grounding for this requirement. Their central finding is that both transparency and human discretion enhance perceived legitimacy of automated decision-making, but that incongruence between citizen preferences and administrative practice signals a misalignment that purely technical transparency cannot resolve (Hillo et al., 2025, p. 1). The democratic implication is stated directly in their study: governing by AI ought also to be responsive to the view of citizens, whose support democratic governing ultimately rests on (Hillo et al., 2025, p. 1). Legitimacy requires not only that citizens can see how decisions are made but that those decisions remain responsive to the citizens who are governed by them. Hillo et al. find that public administrators show a stronger positive response than citizens to both the disclosure of algorithmic processes and to decision-making frameworks that include a human in the final decision (Hillo et al., 2025, p. 10). The asymmetry matters: the addressability gap is deeper for citizens than institutional actors, confirming that the gap cannot be closed by governance arrangements designed primarily around administrative rather than citizen- facing accountability. In this respect, there is a mirror of non-digital addressability. Where there is challenge and no response, trust is eroded. Communication with digital systems is no different, except that those previously working in the non-digital justificatory relationship are removed, with interactions pushed to higher tiers of administration. Toledo confirms that this addressability gap is structural rather than incidental. In traditional administrative systems, decisions are made by identifiable public officials who can be held accountable for their actions. When algorithmic systems influence those decisions, it becomes more difficult to understand how outcomes are produced and the chain of accountability that democratic legitimacy requires is correspondingly weakened (Toledo, 2026, p. 2). Institutional proxies can supplement but cannot substitute for this relationship. An ombudsman acting on the individual's behalf, for example, represents what Diamond and Morlino describe as horizontal accountability: institutional oversight operating at one remove from the individual whose space of reasons is being structured (Diamond and Morlino, 2004, p. 25). This agent is necessary. However, they do not, on their own, discharge the obligation that constitutive power generates. The justification for decisions shaping political agency must be traceable and examinable under a democratic lens. Without a genuine addressee, there is no traceability. Without justificatory addressability, the second condition of the thin standard cannot be met.
The third requirement is minimal co-authorship. If transparency provides a means of determining how a system works and addressability provides somewhere to direct a challenge, co-authorship asks a prior question: do agents have any recognised standing in shaping the conditions of the system itself? Democratic legitimacy, working with Forst, requires that those subject to a normative order are able in some meaningful sense to participate in determining its basic conditions (Forst, 2024, p. 187). This is not a requirement for collective design of every parameter or architectural choice. It is a demand that the architecture governing democratic agency is not constituted entirely without reference to those it governs. Toledo's systematic review identifies five key governance dimensions in the literature on algorithmic accountability in public administration: transparency, explainability, human oversight and administrative responsibility, ethical governance and public trust (Toledo, 2026, pp. 16-17). These dimensions provide a measurable framework that could, in principle, function as a rubric for the ninth dimension in much the same way that V-Dem measures participation: the greater the number of features present and functioning, the stronger the case that the architectural conditions of democratic agency are being met. What Toledo's review also confirms, however, is that existing governance frameworks extend traditional accountability mechanisms to algorithmic systems without yet engaging the prior question of whether citizens have any structural role in constituting the conditions under which those systems operate. The frameworks reviewed address accountability at the level of institutional governance, without resolving how citizens, as agents in the democratic system, can participate in determining the basic conditions of algorithmic governance itself (Toledo, 2026, p. 24). Hillo et al. finding that incongruence between citizen preferences and administrative practice represents a misalignment that technical transparency by itself cannot resolve aligns with Toledo: the legitimacy deficit is not one of information alone but of structural participation (Hillo et al., 2025, p. 1). Both studies confirm the same absence. The existing literature knows what governance mechanisms are needed at the institutional level. It has not yet resolved the prior question of citizen standing in relation to the systems that govern them. The minimum threshold the co-authorship condition sets is thin. It asks only whether citizens have any recognised standing in the processes through which the governing logic of participation is determined, revised or contested. A democracy adjudicating between competing claims presupposes that those making the claims have had some role in constituting the conditions under which their claims can be made. Where that condition is absent, the adjudication is not democratic, however procedurally correct its form. The 300,000 citizens who petitioned against the Palantir contracts in April 2026 illustrate the condition precisely (38 Degrees, 2026a; 38 Degrees, 2026b). They had standing to petition Parliament. They had no standing in relation to the systems themselves: no recognised claim against the architecture, no institutional channel through which a demand for justification could be addressed to those exercising constitutive power over the conditions of their democratic life. Counting, in Dean's sense, without being heard (Dean, 2005, p. 61).12 Taken together, the three requirements specify the minimum structural conditions without which the basic statuses of equality and agency that Diamond and Morlino's framework presupposes cannot function at the architectural level. In the contemporary landscape of 2026, these architectural prerequisites remain systematically absent from both commercial platforms and public administrative systems. Toledo's review of 78 publications confirms this: existing governance frameworks have extended traditional accountability mechanisms to algorithmic systems without yet integrating the structural conditions necessary for citizens to participate in constituting the governance of those systems themselves (Toledo, 2026, p. 24).
Placing the ninth dimension within Diamond and Morlino's framework requires further investigation of accountability and the distinction between vertical and horizontal. Diamond and Morlino define accountability as the obligation of elected political leaders to answer for their political decisions when asked by voters (vertical) or constitutional (horizontal) bodies (Diamond and Morlino, 2004, p. 25). Vertical accountability describes the structured relationship through which citizens hold those who exercise power over them directly to account, operating through electoral and non-electoral channels. Horizontal accountability describes the system of institutional checks through which state actors are held accountable by other state actors: courts, legislatures, ombudsmen and oversight bodies. Platforms exercise a form of power that falls within vertical accountability. They go beyond influencing beliefs or preferences. Rather, they determine the conditions under which individuals become political agents. This is qualitatively different from the influence of traditional media. A newspaper shapes the informational environment. A platform shapes the justificatory conditions within which that information is formed, encountered, assessed and acted upon. The distinction is not one of scale or pervasiveness in the same way that a niche or universal publication is a matter of scale or pervasiveness. It is one of structural function. Where a medium carries political communication, a platform constitutes the conditions of political communication, actively altering the character of public communication itself rather than simply transmitting it (Habermas, 2022, p. 159). This constitutive role is what places platforms within the category of actors toward whom vertical accountability obligations run.
This argument rests on recognising that the form of power platforms exercise—constitutive power over persons as political agents at scale—generates an obligation of justification regardless of the institutional form through which it is exercised. Srivastava identifies platforms as a form of private authority engaged in authoritative decision-making that was previously the prerogative of sovereign states (Srivastava, 2021, p. 990), while Gowder confirms that platforms carry out tasks that are meaningfully similar, in both their social role and the constraints they face, to tasks carried out by states (Gowder, 2023, p. 43). The Forstian argument does not require that platforms be classified as sovereign actors. It requires only that any actor exercising constitutive power over the space of reasons owes justification to those subject to it. Platforms exercise this, and the obligation of justification follows directly from that exercise of power. I identify an imbalance. Platforms exercise constitutive power over individuals as political agents. That power creates a vertical obligation of justification. What exists, however, is largely horizontal: state institutions oversee platforms, courts adjudicate disputes and regulators enforce compliance standards. The gap between the obligation that exists and the accountability mechanisms available to discharge it is not a matter of regulatory ambition or political will. It is structural. Horizontal accountability operates at one remove from the individuals whose space of reasons is being structured. It addresses the platform as a regulated entity with compliance obligations rather than as a constitutive authority with justificatory obligations to those it governs.
The problem is compounded by a second dimension of horizontal infiltration. Horizontal accountability is not only insufficient in relation to platforms, it is subject to the same architectural conditions it is designed to govern. The DWP, the MoD and police forces are horizontal accountability actors whose own operations are governed by algorithmic systems they cannot fully scrutinise, as Stages 1 and 3 of this investigation document. The TAG register identifies 55 active algorithmic systems across UK public administration, the majority operating with limited transparency (Public Law Project, 2024). Courts applying judicial review are restricted to procedural compliance rather than architectural scrutiny, as Thompson and Carlo confirms (R (Thompson and Carlo) v Commissioner of Police of the Metropolis [2026] EWHC 915 (Admin), para 5). In significant respects, architectures have already moved inside the institutions that horizontal accountability relies upon to function.
The European Commission's enforcement programme, under the Digital Services Act, illustrates both the growing reach and the persistent structural limitation of horizontal mechanisms. Since formal proceedings were opened against TikTok in February 2024 (European Commission, 2024), the Commission has moved through advertising transparency, researcher access and, in February 2026, addictive design (European Commission, 2026). The February 2026 preliminary findings are the most architecturally significant yet: for the first time, enforcement action is directed not at what the platform distributes (its outputs) but at how its design shapes user behaviour, specifically the rabbit- hole effect of its recommender systems, infinite scroll, autoplay and push notifications. This represents a meaningful step toward the architectural level. The Commission has recognised that design choices, not only content decisions, carry democratic and social consequences sufficient to attract regulatory intervention. What the February 2026 findings do not yet address is the justificatory standing of individuals in relation to those design choices. Addictive design is engaged as a risk to individual wellbeing, particularly for minors and vulnerable users. The prior question, whether the architecture shaping the conditions of political agency requires justification to those it constitutes as citizens, remains outside the frame. The normative gap I identify is not closed by the February 2026 findings. It is, however, brought into closer proximity to regulatory visibility than at any previous point in the DSA enforcement arc.13
The ninth dimension can now be stated in its final form. Algorithmic Accountability measures the degree to which the systems that constitute individuals as political agents are subject to the justificatory obligations those individuals are owed. It operates at the level of the preconditions of democratic agency: the architectural conditions within which political preferences are formed, through which political claims are made and by which the state's exercise of power over citizens is increasingly mediated. It does not replace the existing eight dimensions. What this new dimension provides is the level of analysis at which their shared presupposition, a background environment within which democratic activity takes place, can itself become an object of democratic assessment. The normative foundation of the dimension rests on Forst's principle of justification, grounded in the conditions of non-domination rather than in a prior metaphysical account of persons. Any agent exercising constitutive power over the space of reasons within which political agency is formed owes justification to those it governs. Platforms, and the algorithmic systems through which public authorities increasingly exercise power over citizens, occupy precisely this position. The obligation they generate cannot be discharged by horizontal accountability mechanisms alone, however sophisticated those mechanisms become. What is required is a form of vertical justificatory accountability that connects those subject to algorithmic systems directly to those who govern them.
The dimension is assessed against a thin normative standard defined by three minimum requirements: architectural transparency sufficient for principled challenge, justificatory addressability providing a genuine locus of accountability and minimal co-authorship ensuring a structural role for individuals in the governance of the conditions of their participation. These requirements are deliberately minimal. They specify only what must be in place before the basic statuses of equality and agency that Diamond and Morlino's framework presupposes can function at the architectural level. They are, however, conditions that do not exist. In 2026, across both commercial platform environments and the algorithmic systems through which public authorities exercise power over citizens, all three remain systematically absent or significantly underdeveloped. The gap the ninth dimension addresses is not an empirical gap within the existing dimensions. It is a structural gap prior to them. Closing it does not require revising the framework's normative commitments, which remain as analytically powerful and empirically measurable as they were in 2004. It requires adding a dimension capable of assessing the conditions that make those commitments realisable in an environment that the original framework was not designed to govern. General Jones in 2004 and 2026 is catchable. The architecture that produces the conditions for General Jones is not, at least not yet, and not without the ninth dimension. That is what Algorithmic Accountability provides: the level of analysis at which the preconditions of democratic life can themselves be examined, contested and held to account.
The underlying principle is not new. Democratic states have long acknowledged that private actors operating within the public sphere generate democratic obligations. In the UK context, this is reflected in media plurality rules and restrictions on ownership, designed to prevent excessive concentration of influence over public political life. The infrastructure through which political communication occurs has, in this sense, never been treated as entirely neutral or purely market-driven. The Telegraph sale, which concluded in April 2026 when Axel Springer acquired the Telegraph Media Group (Nandy, 2026), illustrates both the reach and the limits of existing regulatory principle. The process generated two public interest intervention notices, issued in November 2023 and January 2024 (Secretary of State for Culture, Media and Sport, 2023; 2024), each grounded in the plurality considerations set out in section 58(2A) of the Enterprise Act 2002: the need for accurate presentation of news and free expression of opinion. Ofcom and the Competition and Markets Authority were required to assess market share and consider the broader democratic implications of ownership concentration. The regulatory framework demonstrated the power to block acquisition by a UAE-backed entity on plurality grounds, even at the cost of considerable commercial and diplomatic friction: the offer was higher than the final sale and the UAE are considered allies of the UK (Digital Markets, Competition and Consumers Act 2024; House of Commons Library, 2024). However, that same framework offers no equivalent mechanism for contesting the architectural logic of the platforms that shape what UK citizens encounter as political reality. Foreign-owned recommendation systems operate at a scale, and with a constitutive reach over political agency, that no single newspaper ownership structure could match, and they attract none of the democratic scrutiny the Telegraph process required. The plurality of the information environments those systems create, and the algorithmic processes through which the state itself increasingly makes decisions about citizens, is not subject to equivalent scrutiny. Intervention notices have not been issued, largely because no mechanism exists to trigger one. The principle that control over the public sphere carries democratic responsibility is embedded in existing regulatory practice. What the Telegraph process confirms is that the principle operates only at the level of ownership and editorial identity, not at the architectural level at which the conditions of political agency are increasingly constituted. Where ownership is deemed a threat to national sovereignty or democratic governance, access to UK consumers can be blocked.
The same displacement of principle from architecture is visible in data protection law, where the historical record is instructive. The Data Protection Act 1984 was enacted at the start of significant digital storage becoming possible, reflecting an early recognition that digital systems could be weaponised against individuals and that legislative protection was required. Successive revisions, culminating in the GDPR and its domestic incorporation through the Data Protection Act 2018, have retained and deepened that core commitment to the individual as a data subject with rights over how their personal data is collected, stored and processed. This represents a meaningful protection of informational privacy. However, the framework carries a structural bypass. At the platform interface, GDPR compliance frequently renders the source of data collection opaque rather than transparent, presenting citizens with terms and conditions that nominally satisfy the consent requirement while obscuring what is actually being agreed to. The data harvested under that consent is then aggregated to feed the very algorithmic architecture the regulation does not reach. GDPR regulates the exchange of data, but the exchange it regulates is the mechanism through which the conditions of political agency are shaped, not the protection of those conditions. The legislation addresses the transaction and leaves the architecture untouched.
The Online Safety Act 2023 represents an attempt to address platform power within existing UK regulatory frameworks. The Act introduces a duty of care framework requiring platforms to conduct risk assessments in relation to illegal content and content harmful to children, making Ofcom the regulatory body responsible for overseeing compliance (Online Safety Act 2023a; Ofcom, 2024). However, the risk assessments the Act requires are conducted by the platforms themselves, resulting in the governing logic of the architecture being evaluated by those who designed and profit from it, without any requirement to make that logic externally contestable or independently verifiable (Ofcom, 2023). The Act is directed at identifiable harms at the level of individual pieces of content or categories of users. It does not require platforms to justify the recommendation and curation logic that determines which content reaches which citizens, or to make that logic open to challenge by those subject to it. The House of Commons Science, Innovation and Technology Committee concluded in July 2025 that the online safety regime has major problems and is already out of date, identifying a specific regulatory gap in relation to recommendation algorithms and finding that the OSA fails to address the algorithmic amplification of content in ways that endanger public safety (House of Commons Science, Innovation and Technology Committee, 2025). What the Online Safety Act regulates is the surface at which harm becomes legible. The architectural conditions under which citizens are constituted as political agents, and under which some content reaches citizens, remain outside the accountability relationship the Act creates.
Most recent announcements regarding the Children's Wellbeing and Schools Act 2026 extends the OSA's reach by granting the Secretary of State powers to restrict children's access to specified platforms and features (Children's Wellbeing and Schools Act 2026). Announcements in June 2026 of a ban on social media for under-16s from Spring 2027 represents the most direct intervention yet in the relationship between citizens and platform architecture (HM Government, 2026). Yet the concern driving these measures remains one of access and content exposure rather than architectural accountability. The question being asked is which citizens should be permitted to enter the architecture, not whether the architecture itself is subject to democratic justification. Harmful algorithmic recommendations are identified as a concern in Ofcom and ICO enforcement letters to platforms (Ofcom and ICO, 2026), but the remedy proposed is restriction of access, not contestation of the governing logic. This is the most ambitious child protection intervention in a generation, but remains directed at the phenomenal surface. The architectural level at which the conditions of political agency are constituted remains unaddressed.
The Digital Services Act represents the European Union's most ambitious attempt to reach platform power at a structural level. Its provisions on very large online platforms impose risk assessment obligations under Article 34 and mitigation duties under Article 35, with the European Commission holding enforcement powers that can reach up to six per cent of a platform's global annual turnover (Regulation (EU) 2022/2065). The Commission's preliminary findings of 6 February 2026 against TikTok mark a significant regulatory moment. For the first time, enforcement action focused not on illegal content, data protection or competition, but on the addictive design features of the platform architecture itself, including infinite scroll, autoplay, push notifications and highly personalised recommender systems (European Commission, 2026), the same features that informed the UK government's June 2026 announcements on children and social media. The Commission found that TikTok had failed to adequately assess how those features harmed the physical and mental wellbeing of its users, particularly minors and vulnerable adults, and had disregarded indicators of compulsive use including the time spent on the platform at night. Even here, however, the accountability relationship runs between the platform and the regulator, assessed by reference to risk to individual wellbeing rather than to the democratic conditions of political agency. The DSA reaches the architecture. It does not yet reach the question of whether that architecture is subject to the justificatory obligations it owes to those whose conditions of political reasoning it constitutes.
Canada provides a different but equally instructive illustration of the limits of content-level regulation. The Online News Act 2023 (Bill C-18) attempted to require digital platforms to compensate Canadian news publishers for content distributed through their services. Meta's response was immediate and architectural: on the day the Act received Royal Assent, the company confirmed it would end news availability on Facebook and Instagram for all Canadian users, implementing the block in August 2023 (Meta, 2023). The platform did not contest the principle. It reconfigured its architecture to exclude the regulated content entirely, rendering the regulation effectively irrelevant in relation to the democratic information environment it was designed to protect. The state had attempted to reach platform behaviour at the level of content distribution. The platform responded by exercising its architectural power at a level the legislation could not reach. As of June 2026, nearly three years after the block was implemented, news has not been restored. The Canadian government remains in negotiations with Meta to persuade the platform to reverse an architectural decision it made unilaterally, with the United States having identified the Online News Act as a trade irritant in bilateral negotiations—a key feature relied on by major platforms when faced with government-scale attempts at regulation. The regulator is negotiating with the regulated. Canada's parallel attempt to introduce a duty-of-care framework through the Online Harms Act (Bill C-63) compounded the picture. Modelled on the UK's Online Safety Act, the Bill defined harms as seven specific categories of content, most of which were already illegal, and included a provision requiring removal within twenty-four hours where directed by the Canadian government. This left platforms otherwise free to determine whether content meeting those definitions should be removed (Department of Canadian Heritage, 2024). Even had it passed, the self-regulation status quo would have remained essentially intact at the architectural level (Centre for International Governance Innovation, 2024). The Bill fell twice on the order paper, most recently in January 2025 when Parliament was prorogued following the Prime Minister's resignation, its failure driven by political circumstance rather than parliamentary opposition to its intent (TechPolicy.Press, 2025a). The Canadian case demonstrates two things simultaneously: that platforms will deploy their architectural power to neutralise regulation directed at the content level and that even the most ambitious content-level frameworks being attempted in the developed West do not reach the architectural conditions under which the democratic information environment is constituted.
Across these Western frameworks a consistent pattern emerges. The principle that private actors exercising power over the public sphere attract democratic obligations is not only recognised but actively enforced at the level of ownership, content and data exchange. The Telegraph process demonstrates that the state will act on that principle even at a considerable economic and diplomatic cost. I argue that what none of these frameworks settle is the architectural-level problem: the level at which the conditions under which individuals encounter political reality are constituted before any observable democratic activity takes place. The displacement is not accidental. It reflects the limits of regulatory categories developed before platform architecture existed as a political phenomenon.
The regulatory pattern identified in the developed West is not confined to contexts where sophisticated legal frameworks exist. Globally, democratic states have engaged with platform power in ways that range from the exceptionally stringent to the formative. What is consistent across that range is not the quantity or intensity of regulation but its object: in every case, regulatory effort is directed at the visible outputs of platform power, at content, compliance, data exchange and user identification. In common with Western democracies, the architectural conditions under which political agency is constituted remain unaddressed. The gap my ninth dimension identifies is not a function of regulatory capacity. It is a function of regulatory object.
India demonstrates this most sharply: a framework more demanding in several respects than anything the developed West has produced but similarly empty on the architectural level. India provides the most developed example and in some respects the most instructive one. The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules 2021, issued under the Information Technology Act 2000, introduced substantive obligations on significant social media intermediaries that have been progressively tightened through four rounds of amendment, most recently in February 2026 (Ministry of Electronics and Information Technology, 2021; 2026).14 In several of its provisions India's framework is more stringent than anything the OSA, GDPR or the DSA requires. The intensity of regulatory effort is not in question. What the framework does not do, in any of its iterations, is address the recommendation or curation logic that determines which content reaches which citizens and under what conditions of political salience. India demonstrates that a state can construct an exceptionally demanding regulatory apparatus and still not engage with the architectural level, because all of that intensity is directed at what platforms distribute, who produced it and how quickly it can be removed. The conditions under which citizens are constituted as political agents remain entirely outside the accountability relationship the rules create.
Brazil presents a different but equally instructive case. Brazil's PL 2630, introduced in 2020 and debated across three legislative sessions, is a focused attempt by a democracy to address platform power at the architectural level. Explicitly modelled on the DSA, the bill proposed algorithmic transparency requirements, limits on automated content sharing and independent oversight mechanisms that would have gone beyond mere content removal obligations (Tomaz, 2023, pp. 255- 257). Its ambition was precisely what my ninth dimension identifies as necessary: it attempted to make the governing logic of recommendation systems externally contestable rather than trusting regulation of the content those systems amplify. Tomaz concludes, however, that the bill follows a general pattern of platform regulation in that “measures that fail to alter the fundamental business model and dynamics of the firms themselves will leave in place the incentives that drive the problematic practices of surveillance and data mining and negative forms of misinformation and amplification” (Tomaz, 2023, p. 264). The bill was defeated. The way in which it was defeated is analytically significant. Research by NetLab at the Federal University of Rio de Janeiro documented that major platforms including Google and Meta ran campaigns against PL 2630 using the very algorithmic amplification mechanisms the bill was designed to address, promoting content against the legislation through paid promotion and recommendation systems in ways that circumvented their own advertising policies (Global Freedom of Expression, Columbia University, 2023). The architecture was deployed to defeat the legislation that would have made it accountable. The bill died on the order paper at the end of 2024 with no immediate prospect of revival (Internet Society, 2025).
The episode also exposes the limits of the quasi-state governance structures that platforms have developed as alternatives to external regulation. Meta's Oversight Board, established in 2020 as an independent body to review content moderation decisions, is the most prominent example of this model (Meta, 2020). It positions the platform as a self-governing entity capable of holding its own architectural decisions to account. Yet the Oversight Board had no role in, and no jurisdiction over, the algorithmic campaign Meta ran against PL 2630. The internal accountability structure was entirely irrelevant to the exercise of architectural power that defeated the legislation that would have subjected that power to external justificatory obligations. This is precisely the limitation Forst's principle identifies: justificatory obligations cannot be discharged by the entity that holds the power, directed at itself, on its own terms. The Brazil case makes that limitation visible not as a theoretical proposition but as a political event with measurable democratic consequences.
Indonesia and South Africa serve to complete an emerging democracies picture. Together they illustrate the range of positions within which the same fundamental gap appears. Indonesia's Electronic Information and Transactions Law, most recently amended as Law No. 1 of 2024, combined with the Personal Data Protection Law of 2022, creates a layered regulatory framework addressing electronic transactions, content obligations and data protection that is among the most developed in Southeast Asia (Republic of Indonesia, 2022; 2024). The 2024 amendment broadened government oversight powers over electronic system providers and introduced child protection measures in digital environments. Like India's IT Rules, however, the framework is directed entirely at content categories, data exchange and compliance obligations. The recommendation and curation logic through which Indonesian citizens encounter their political environment remains outside the accountability relationship the law creates.
South Africa represents the earliest stage of the regulatory journey among the four cases examined here. The Protection of Personal Information Act 2013, fully enforceable since July 2021, provides a GDPR-adjacent data protection framework with meaningful enforcement powers, including fines and criminal penalties for serious breaches (Protection of Personal Information Act 2013, s. 4). Section 71 of POPIA addresses automated decision making, providing that a data subject may not be subject to a decision which results in legal consequences or which affects them to a substantial degree, based solely on automated processing of personal information intended to provide a profile of that person (Protection of Personal Information Act 2013, s. 71(1)). This is significant: there must be an individual in the justification process for that process to be valid. Where exceptions apply, the responsible party must provide the data subject with sufficient information about the underlying logic of the automated processing to enable them to make representations (Protection of Personal Information Act 2013, s. 71(3)(b)). Again, this is an attempt to enforce a justificatory process. This provision requires disclosure of underlying algorithmic logic, not merely of the decision itself. Yet the provision applies only where a decision has already been taken and its effects are already legally significant. It does not reach the conditions under which citizens are constituted as political agents before any such decision occurs, and it does not create a contestable accountability relationship between citizens and the recommendation systems that shape their political environment. The Draft National AI Policy opened for public comment in April 2026 signals growing governmental awareness of algorithmic risk, identifying algorithmic bias, accountability in autonomous decision- making and transparency as priorities for future legislation (Department of Communications and Digital Technologies, South Africa, 2026). South Africa is building the foundations for architectural accountability.
Taken together, these four democratic cases establish that the pattern identified earlier is not a product of the specific regulatory culture of the developed West. It is a structural feature of how democratic states engage with platform power regardless of their institutional context. The most ambitious attempt, Brazil's PL 2630, failed in part because the architecture it sought to regulate was deployed against it. The most stringent emerging democracy framework, India's IT Rules, has been amended repeatedly without ever reaching the architectural level. Indonesia's layered framework follows the same content-level logic despite its sophistication. South Africa is building the foundations of a regulatory response that has not yet arrived at the architectural question. The democratic challenge posed by platform architecture is global. The regulatory gap the ninth dimension identifies is equally global. Its consequences, however, are not uniform: they are severest where the compensating mechanisms of institutional oversight, judicial independence and civil society capacity are weakest. Such weaknesses provide opportunities for platforms to weaponise their architecture against state intervention.
Alongside the issue of states wrestling control from algorithmic systems is the ability of states to engage in undemocratic activity. The most severe instances of architectural power over democratic agency are found in hybrid and autocratising states where platform architecture interacts with, or is actively aligned with, state-directed political closure. Here the absence of architectural accountability is not just a regulatory problem but a condition that accelerates democratic collapse. I use three cases to illustrate different modalities of the same fundamental problem: architecture operating without justificatory obligation in contexts where the democratic stakes are highest.
Myanmar provides one of the most extensively documented cases for the ninth dimension's argument. The UN Independent International Fact-Finding Mission on Myanmar concluded in its September 2018 report that Facebook had been a useful instrument for vilifying the Rohingya, noting that for most users in Myanmar, Facebook was the internet (UN Human Rights Council, 2018). The Mission found that the role of social media was a significant element in the atrocities perpetrated against the Rohingya and called for senior military officials to be investigated and prosecuted for war crimes, crimes against humanity and genocide. Amnesty International's 2022 report went further, establishing through analysis of internal Meta documents that the platform's algorithms proactively amplified and promoted content inciting violence, hatred and discrimination against the Rohingya (Amnesty International, 2022). The architectural logic went beyond failure to remove harmful content (an act of omission). It actively constituted the political environment within which genocide became possible. An illustration of this dynamic occurred when civil society activists launched the Panzagar, or flower speech, initiative in 2014, encouraging users to post counter-speech stickers in response to anti- Rohingya content. Facebook's engagement-based algorithms interpreted the stickers as a signal that users were enjoying the posts and further amplified the hate content they were designed to counter (Amnesty International, 2022). The architecture was constitutive of the conditions under which violence reached political scale. There was no accountability relationship between those whose democratic and human conditions were being shaped and destroyed and the systems doing the shaping. The architecture was not contested because it was not recognised as contestable.
Ethiopia illustrates the same pattern in a conflict context where the architectural failure is documented and its consequences traceable to an identifiable individual. The Tigray conflict was characterised from its outset by the use of Facebook to spread hate speech and incite ethnic violence (TechPolicy.Press, 2025b). Frances Haugen told the US Congress in 2021 that Facebook was fuelling ethnic violence in Ethiopia's civil war. Amnesty International's October 2023 report documented how Facebook's algorithmic systems supercharged the spread of harmful rhetoric targeting the Tigrayan community, while the platform's content moderation systems failed to detect and respond appropriately to such content (Amnesty International, 2023). The report established that these architectural failures contributed directly to the killing of Professor Meareg Amare, a Tigrayan chemistry professor at Bahir Dar University, who was targeted in Facebook posts on 3 November 2021 containing identity and location details, with false claims that he was a supporter of the Tigrayan People's Liberation Front. His son told Amnesty: “I knew it would be a death sentence for my father” (Amnesty International, 2023). The posts were algorithmically amplified before any moderation response was triggered. The Meta Oversight Board's engagement with an Ethiopian case in October 2022, in which it upheld the removal of a post from the Tigray Regional State's Communication Affairs Bureau, illustrated simultaneously the potential and the limits of internal governance: the Board's jurisdiction extended to individual content decisions, not to the algorithmic architecture that determined the reach and amplification of content across the platform as a whole (Global Freedom of Expression, Columbia University, 2022). The governing logic of the system remained outside any accountability relationship.
Nigeria presents a third and differently structured modality. Where Myanmar and Ethiopia illustrate architecture operating without accountability in contexts of communal and ethnic violence, Nigeria illustrates what happens when a platform applies its own content rules to a head of state and the state responds with architectural suppression. On 5 June 2021, Nigerian authorities directed internet service providers to block access to Twitter after the platform deleted a tweet by President Muhammadu Buhari for violating its rules on hateful conduct. The government's justification was the platform's persistent use for activities capable of undermining Nigeria's corporate existence—a formulation sufficiently vague to encompass any political content the government found inconvenient (Electronic Frontier Foundation, 2022). The suspension lasted 222 days and was lifted in January 2022 after Twitter agreed to register a legal entity in Nigeria and comply with local content requirements. The ECOWAS Community Court of Justice subsequently ruled the suspension unlawful as a violation of freedom of expression under the African Charter on Human and Peoples' Rights and the International Covenant on Civil and Political Rights (SERAP and Others v Federal Republic of Nigeria, ECW/CCJ/APP/23/21, 2022). The Nigeria case is significant for my argument in a specific way: it shows that the absence of architectural accountability runs in both directions. Platforms can exercise architectural power over citizens without justificatory obligation and states can exercise suppressive power over platforms without justificatory obligation to the citizens whose democratic information environment is thereby destroyed. In neither direction does the accountability relationship the ninth dimension requires exist.
Across these three cases, a pattern distinct from those identified in the Developed West and Emerging Democracies sections emerges. In the former, architectural accountability is displaced by regulatory categories that address the phenomenal surface. In the latter, the same displacement operates with fewer compensating mechanisms. In states where democracy is under pressure or actively retreating, the absence of architectural accountability is not a displacement but a condition of political closure. The architecture operates, is suppressed or weaponised and in each case the citizens whose conditions of political agency are at stake have no standing to contest it. The ninth dimension identifies the gap that exists in sophisticated regulatory frameworks and the democratic floor below which the absence of any such framework allows societies to fall.
Platform-side regulatory obligations, even where they approach the architectural level, are necessary but not sufficient conditions for the kind of accountability the ninth dimension requires. Forst's principle of justification requires there is somewhere to direct a justificatory claim and that those making the claim have the capacity to form and direct it. A citizen who has never been taught to recognise algorithmic curation as a condition of their own reasoning is doubly disadvantaged: they cannot identify the structural condition and they cannot name an addressee for a claim about it. The question of citizen-side capacity is a precondition for the vertical justificatory accountability the ninth dimension requires to be meaningful in practice. The UK's approach to this precondition is partial but instructive. The GCSE syllabus, such as the OCR Computer Science specification, requires all students to engage with the ethical, legal, cultural and environmental impacts of digital technology as a mandatory assessed component, including the purposes of data protection legislation and the impacts of digital technology on wider society (OCR, 2024a). This represents a baseline state recognition that citizens require some capacity to engage critically with the digital conditions of their own lives. At A-Level, the OCR Computer Science specification (H446) goes further, requiring substantive engagement with automated decision-making, artificial intelligence, algorithmic bias and accountability and the question of who bears responsibility when an algorithm makes a consequential decision (OCR, 2024b). These provisions embed a justificatory logic at the level of secondary education: students are expected to understand not only what digital systems do but who is responsible for what they do and on what grounds that responsibility can be challenged.
Statutory grounding for this educational function sits within Ofcom's media literacy duties under section 11 of the Communications Act 2003, which requires Ofcom to promote and facilitate the development of media literacy among members of the public (Communications Act 2003, s. 11). The
Online Safety Act 2023 extended and strengthened these duties, placing specific obligations on Ofcom to produce a media literacy strategy and to take steps to improve public understanding of the nature and impact of regulated content (Online Safety Act 2023b, s. 143). The statutory embedding of media literacy distinguishes the UK's approach from purely voluntary or market-led educational initiatives and gives it a normative significance for the ninth dimension's argument. What these educational and statutory provisions do not yet do is address the architectural level at which the ninth dimension is directed. The OCR specifications, for example, teach students to recognise algorithmic bias and to understand data protection law. They do not yet give students, or the citizens they will become, the structural standing to contest the governing logic of the systems that curate their political environment, or to direct justificatory claims at the persons or bodies responsible for those systems. Ofcom's media literacy strategy addresses how citizens engage with content, but does not address how citizens might contest the architectural conditions under which that content is selected and prioritised. The citizen-side precondition is being partially addressed. The architecture-side obligation identified by the ninth dimension remains largely unmet. The gap between what existing provision supplies and what architectural accountability would require on the citizen side is precisely what the research agenda developed next addresses.
Across all four registers I have examined, a consistent pattern emerges. The principle that private actors exercising power over the public sphere attract democratic obligations is not new. It is embedded in media plurality law, as the Telegraph process demonstrates; in data protection regulation, as GDPR and POPIA illustrate; in content governance frameworks, as the OSA, the DSA and the IT Rules 2021 confirm; and in statutory media literacy duties, as Ofcom's obligations under the Communications Act 2003 and the Online Safety Act 2023 establish. What has not been achieved, in any democratic context, is the application of that principle at the architectural level: the level at which the conditions of political agency are constituted before any observable democratic activity takes place.
In the developed West, the principle is recognised and enforced at the level of ownership, content and data exchange, but consistently displaced from the architectural level. The Telegraph process evidences that the state will act on plurality grounds even at economic and diplomatic cost, yet offers no equivalent mechanism for contesting the governing logic of recommendation systems that shape political reality at far greater scale. GDPR regulates the exchange of data but leaves intact the architectural logic that determines what individuals encounter as political reality before any data exchange occurs. The OSA and DSA, the most ambitious frameworks yet produced, reach further than anything before them but still fall short: the TikTok preliminary findings of February 2026 represent the furthest point any regulatory framework has reached and the accountability relationship they create runs between platform and regulator rather than between platform and citizen. Canada demonstrates that platforms will deploy their architectural power to neutralise legislation directed at the content level, while Brazil shows that even legislation explicitly targeting the architectural level can be defeated by the architecture it seeks to regulate.
In emerging democracies, the same displacement operates with fewer compensating mechanisms and in some cases with greater regulatory intensity directed at the wrong object. India's framework is more stringent, in many respects, than anything the developed West has produced, but equally silent on the architectural level. Brazil came closest to addressing the architectural question directly only to be defeated by the architecture it sought to regulate. Indonesia and South Africa illustrate respectively a sophisticated content-level framework and the earliest stage of a regulatory journey that has not yet arrived at the architectural question. In states where democracy is under pressure or retreating, the absence of architectural accountability is not a displacement but a condition of political closure. Myanmar demonstrates that platform architecture can actively constitute the conditions under which genocide becomes possible. Ethiopia demonstrates that those consequences can be traced to identifiable individuals. Nigeria demonstrates that the accountability deficit runs in both directions: platforms exercise architectural power over citizens without justificatory obligation and states suppress platforms without justificatory obligation to the citizens whose democratic information environment is thereby destroyed.
The citizen-side dimension, examined through the UK's educational and statutory provision, confirms that the gap is not only a platform-side or state-side failure. Even where states are building the citizen- side preconditions for justificatory accountability through curricula and media literacy duties, those provisions address how citizens engage with content rather than how they might contest the architectural conditions under which that content reaches them. The ninth dimension identifies this consistent gap across varying states of government and, crucially, how education can be used to tackle some of the epistemic requirements. It does not offer a specific regulatory solution. That is the task of the research agenda the final stage develops. What I have established through this analysis is that the gap is real and that it is structural rather than incidental, continuing to cause deadly harms. The displacement of democratic principle from the architectural level is not a recent failure. It is a structural feature of how states have engaged with private power over the public sphere, now made more consequential by the scale and reach of the architecture through which that power operates.
The central claim of my paper is that platform architectures exercise a form of constitutive power over individuals as political agents. That power is noumenal in Forst's sense: it operates by shaping, occupying and foreclosing the space of reasons. The obligation of justification this generates is not met by existing accountability structures. Algorithmic systems corrupt both directions of accountability simultaneously. At the horizontal level, they have embedded themselves within the institutions designed to provide oversight, so that the DWP, the MoD and the police are themselves governed by architectures they cannot fully scrutinise. At the vertical level, they pre-structure the preferences and agency that citizens bring to the relationship with government before that relationship begins. The justificatory connection that democratic legitimacy requires, running directly between those subject to these systems and those who govern them, does not yet exist in any systematic form.
What follows from this is a shift in how democratic quality must be assessed. The existing eight dimensions remain capable of measuring what they were designed to measure. They operate at the level of observable outputs. Without a dimension capable of assessing the conditions that produce those outputs, there is a structural risk: high scores across participation, responsiveness, freedom, and formal legality mask a deeper and growing deficit in the conditions under which democratic agency is possible. The ninth dimension does not replace the original framework. It addresses the level of analysis the original framework presupposes but cannot itself examine.
The paradox this creates is visible in contemporary democratic life. Participation has never been more formally extensive. That formal visibility does not correspond to an increase in justificatory agency. The conditions under which participation carries democratic weight are being systematically narrowed at the same time as the phenomenal indicators of participation are expanding. In Dean's terms, citizens are engaged in the fantasy of participation, contributing to the infostream in the belief that it registers (Dean, 2005, p. 60), while the architectural conditions that would give those contributions democratic weight are being hollowed out. The argument therefore is not that Algorithmic Accountability would improve Diamond and Morlino's framework, but that it has become necessary if the framework is to continue measuring democratic quality under contemporary technological conditions. My investigation contributes a conceptual diagnosis and a normative extension to one of the most widely used frameworks for assessing democratic quality.
Three tasks for future development are generated from my argument. Together they constitute a research agenda for operationalising the ninth dimension: moving from the normative and theoretical account offered here toward the empirical and institutional frameworks that would allow Algorithmic Accountability to function as a measurable dimension of democratic quality alongside Diamond and Morlino's original eight.
The first task is conceptual. The thin standard set out in Stage 4 specifies the minimum conditions for justificatory agency at the architectural level: transparency, addressability and minimal co-authorship. The task is to translate these requirements into operational criteria that could function across different platform environments, regulatory contexts and institutional settings: the horizontal and vertical dimensions of political activity. Toledo's review of governance dimensions in algorithmic accountability in public administration offers a starting point: transparency, explainability, human oversight and administrative responsibility, ethical governance and public trust (Toledo, 2026, pp. 16– 17). These dimensions provide a measurable rubric that could serve the ninth dimension in much the way that V-Dem serves participation. The greater the number of features present and functioning, the stronger the case that the architectural conditions of democratic agency are being met. The conceptual task is to refine and align these criteria to the justificatory standard rather than simply to governance efficiency and to establish thresholds at which the thin standard can be said to have been met. The thin standard is deliberately minimal, leaving open the question of what full democratic accountability for algorithmic systems would look like. That question need not be resolved here, but the conceptual task is to ensure it becomes askable in precise rather than general terms.
The second task is empirical. The dimension requires testing against specific cases across contrasting regulatory environments. The EU represents a supranational rights-architecture approach. The DSA constructs accountability through horizontal legislative frameworks, operating at one remove from individuals through institutional enforcement, researcher access requirements and due process obligations. Its enforcement against TikTok illustrates the reach and the structural limitation of this approach. As the enforcement landscape from 2024 to 2026 shows, horizontal mechanisms can be extended progressively towards the architectural level. They do so through institutional proxies rather than through any mechanism by which individuals themselves could contest the systems that constitute them as political agents. Whether the DSA framework meets the thin standard of the ninth dimension, across any of its three requirements, is an empirical question to be addressed, and doing so would be the first test of whether the thin standard can generate measurable criteria in practice.
India offers a structurally contrasting case. India's Information Technology Rules represent an approach shaped by conditions of democratic scale, platform dependence and uneven digital infrastructure access, rather than by the rights-architecture logic of the EU. The 2025 proportionality requirements and the requirement for senior-level authorisation for content removal orders introduce accountability mechanisms that operate through administrative hierarchy rather than supranational enforcement. The February 2026 mandatory labelling and metadata embedding requirements for AI- generated content engage, at the level of traceability, precisely the problem of transparency and addressability that the ninth dimension identifies. Comparing these two frameworks against the thin standard would allow the research agenda to assess how different regulatory traditions approach the conditions of algorithmic accountability. A framework designed to measure democratic quality across different contexts must be capable of operating across those contexts. The EU/India comparison is well suited to testing that capacity.
Beyond regulatory comparison, empirical work is required at the level of specific systems. Recommender systems, content curation practices and algorithmic identification technologies each instantiate the architectural conditions the ninth dimension addresses in different ways and with different measurable profiles. The Post Office Horizon case is instructive here as a limit condition. Though not a platform in the digital public sphere sense, Horizon illustrates with particular clarity what noumenal erosion looks like when it reaches its logical end: a system whose governing logic was protected by institutional and legal structures to the point where citizens' demands for justification were not merely unanswered but structurally invalidated. The subpostmasters were counted and wrongly prosecuted without being heard. Their experience establishes that the structural conditions the ninth dimension addresses are not confined to social media platforms. They appear in any algorithmically mediated system through which public authority is exercised over citizens and whose governing logic is closed to the justificatory demands of those it governs. Empirical research should therefore engage both commercial platform environments and public administrative systems, adopting the Horizon case as the indicator of a structural condition that the ninth dimension must be capable of detecting wherever it operates.
The primary methodological obstacle for this research agenda is the same condition the ninth dimension identifies. The governing logic of the systems in question is, in most cases, withheld under proprietary exemptions or freedom of information exclusions that place it beyond independent scrutiny. The EU's enforcement proceedings against TikTok illustrate the scale of that obstacle at the regulatory level. Where the highly resourced and legally empowered democratic institution attempting to reach platform architecture faces sustained resistance, individual researchers face a considerably higher one. That inaccessibility is not incidental to the empirical task. It is itself evidence of the transparency deficit the thin standard requires future research to measure. It also shapes the methods available. Regulatory disclosures, published audits, court documents, parliamentary committee evidence and freedom of information responses, however partial, constitute the accessible data layer through which the architectural conditions the ninth dimension addresses can be examined. The Horizon case was reconstructed through those sources. They are not a substitute for direct access to governing logic, but they do establish whether the conditions of the thin standard are present.
The third task is institutional. If platforms and public algorithmic systems exercise a form of power that generates obligations of justification to those subject to them, what institutional mechanisms are needed to make vertical accountability a structural reality rather than a normative aspiration? This is a movement from problem diagnosis to solution design, from a thin normative standard to the thick institutional question of who should receive justificatory demands, through what channels those demands are directed and with what authority to compel a substantive response rather than a procedural one. Existing mechanisms do not meet that standard. Courts applying judicial review are restricted to procedural compliance rather than architectural scrutiny, as Thompson and Carlo confirm. Regulatory enforcement operates after the architectural condition has already operated, addressing rather than preventing the harm. Ombudsmen and oversight bodies can investigate and report but cannot constitute a genuine justificatory addressee in Forst's sense. In each case the accountability relationship runs between institutions rather than between the individual and the system that constitutes them as a political agent. What would it mean for an individual to direct a demand for justification at the architecture itself? The institutional task is to design arrangements capable of answering that question: mechanisms through which such a demand can reach the governing logic of the system, be received as binding and compel a response with genuine architectural force.
The stakes of my investigation extend beyond the question of whether Diamond and Morlino's framework requires an additional dimension. It does. Our digital landscape raises the broader question of whether democratic analysis has kept pace with the structural transformation of the environments within which democratic life takes place. It has not. The V-Dem Democracy Report 2026 records what it describes as an unprecedented democratic recession. Within a single year the United States's Liberal Democracy index score fell by 24 percent, with freedom of expression reaching its lowest recorded level since the end of the Second World War (Nord et al., 2026, p. 8). These are outputs: measurable indicators of democratic deterioration registered by the dimensions the existing framework provides. What those indicators do not show is the structures that produce them or the level at which the democratic harm originates. High formal capacity can coexist with the progressive architectural foreclosure of justificatory agency. The formal indicators and the substantive conditions can diverge. When they do, existing frameworks will continue to measure what is happening without being able to reach the level at which it is being produced. This is the condition my ninth dimension addresses. Directing analytical attention to the architecturally mediated conditions of democratic life provides a framework through which existing democratic practices and future developments can be more systematically evaluated. Without this framework, democratic analysis will continue to focus on what can be measured rather than on what makes those measurements democratically meaningful. Participation, responsiveness and the rule of law will be recorded and compared while the conditions under which participation carries justificatory weight remain outside the frame. The challenge posed by algorithmic systems is therefore not simply one more problem for democratic governance. It is a challenge to the conditions under which democratic governance becomes possible. Until those conditions themselves become objects of democratic scrutiny, measures of democratic quality will remain incomplete.