Introduction
Automated systems have become integral to the everyday operations of the modern state. From algorithmic tools that assist in welfare allocation, policing, and immigration screening, to platforms that shape public-sector management and citizen engagement, these technologies promise efficiency, objectivity, and scalability. But they also raise urgent questions about what it means to be human in a system that no longer simply serves us but often acts upon us.
Decisions once made through human judgment are increasingly filtered through data-driven processes, where bureaucratic discretion gives way to algorithmic classification and legal entitlements become mediated by software. As state power operates through these automated systems, individual rights, human dignity and autonomy, and procedural safeguards face unprecedented challenges, raising critical questions about accountability, procedural justice, fairness, and the rule of law.
This Blog Post Symposium takes up these questions under the title ‘Humanity in the Automated State’. It accompanies a lecture series of the same name held at Leiden University throughout the 2025/2026 academic year, that brought together seven leading scholars from law, management, public administration, and computer science to examine how the automation of state authority affects fundamental aspects of the human experience: being recognized as individuals rather than data points, participating meaningfully in decisions that affect us, protecting those made vulnerable by algorithmic bias, and maintaining space for resistance and appeal. These lectures traced four thematic lines: the meaning and limits of human oversight; the legal and institutional mechanisms through which accountability can be established; the prospects for justice and legal protection in algorithmically governed systems; and what automation does to human dignity, autonomy, and the conditions of a fully human life. The fourteen contributions gathered in this Symposium take up and extend each of these themes in turn.
Human Oversight: What It Means and Whether It Works
Human oversight has emerged as the central governance response to the risks of automated decision-making. Yet the concept raises as many questions as it answers: what does meaningful oversight require, who is the human it invokes, and can it deliver what it promises?
In the lecture series, this theme was explored most directly by two speakers. Aya Rizk (Linköping University) examined how automation reshapes the decision space between public officials and citizens, emphasizing that meaningful human involvement requires attention to the social and institutional conditions in which automated decisions are embedded. Ida Koivisto (University of Helsinki) pushed further, challenging the figure of the human that technology regulation constructs and arguing that legal requirements for human oversight invoke an abstract entity whose relationship to actual human beings is far from clear.
Three contributions in this Symposium pick up these questions.
Jake Goldenfein and Connal Parsley challenge what they call the ‘human strategy’: the reflexive turn to human agency as a corrective to automated administration. They argue that this strategy not only misunderstands the material realities of algorithmically mediated decision-making, but risks foreclosing more fundamental questions about whether the concept of ‘decision’ itself remains a meaningful or adequate category for governing the automated state. Isabella Banks draws on relational sociology and organizational studies to argue that the quality of human oversight cannot be reduced to the design features of a human-machine interface. Introducing a relational theory of human oversight, she shows that effective oversight depends on the human relationships that exist around the loop: between professional AI users, between users and developers, and between users and those affected by automated decisions. Johann Laux examines how the rise of agentic AI systems is transforming the role of the civil servant and the epistemic relationship between the state and its citizens. Drawing on James C. Scott’s concept of ‘seeing like a state’, he argues that automation is shifting bureaucrats from constitutive participants in decisions to corrective monitors of autonomous systems, a transformation with profound implications for what it means to govern and to be governed.
Accountability and Control Mechanisms
The second theme turns to the concrete legal and institutional tools through which accountability over automated systems can be established and maintained. As AI is adopted across public administration, the question is not only whether accountability mechanisms exist, but whether they are fit for purpose, that is, capable of reaching systems whose opacity, speed, and complexity routinely outpace the institutions designed to scrutinize them.
In the lecture series, Madalina Busuioc (VU Amsterdam) argued that accountability cannot be retrofitted after deployment but must be built into the conditions of AI adoption from the outset. In her closing lecture, Natali Helberger (University of Amsterdam) showed how engaging with citizens’ lived experiences through participatory risk assessment offers an alternative to the metrics-based approaches that currently dominate AI governance.
Four contributions in this Symposium develop these concerns further. Robbert Bruggeman examines the transparency challenges posed by rule-based automated decision-making systems in public administration, illustrating these through landmark court rulings from the Netherlands and Spain. He argues that knowledge modelling, representing legal rules in human- and machine-readable form, offers a promising method for making algorithmic systems legible and contestable. Michael Sierra analyses the role of notified bodies under the EU AI Act, third-party conformity assessment institutions tasked with auditing high-risk AI systems. He argues that these bodies should be understood as regulatory intermediaries whose institutional design, independence, and epistemic orientation will determine whether algorithmic auditing becomes a meaningful safeguard or a form of regulatory theatre. Alex Chanhou Lou argues that the legality of AI training data is no longer a private copyright matter but a public-law problem that public buyers must address through procurement. Drawing on US fair use doctrine and emerging litigation, he shows how courts construct ‘market evidence’ that public procurers can translate into enforceable accountability controls. Matias Mascitti proposes a constitutional strategy for overseeing automated administrative decision-making across its full lifecycle, from design through deployment, monitoring, contestation, and revision. Centred on the concept of the ‘script’ and an institutional proposal he calls the Algorithmic Center, the contribution argues that effective constitutional oversight requires coordinated institutional design rather than piecemeal legal intervention.
Justice, Courts, and Legal Protection
The third theme examines what algorithmic governance means for those who seek legal protection against it: in courts, in dispute resolution mechanisms, and in the criminal justice system. Access to justice has always been uneven. Algorithmic governance thus adds new layers of complexity to an already difficult terrain.
In the lecture series, Mengchen Dong (Max Planck Institute for Human Development) presented empirical evidence that public attitudes toward AI in welfare allocation mask a false consensus: non-claimants systematically misread the preferences of those actually affected, with significant consequences for how AI systems are designed and justified.
Four contributions in this Symposium extend this inquiry. Elke H. Olthuis and Anna van Duin bring an empirical, citizen-centred perspective to access to digital justice, reporting findings from a mixed-methods study with Dutch citizens. They show that citizens are more articulate about the trade-offs of digital dispute resolution than is often assumed, and that procedural fairness, particularly the sense of being heard, remains central to their expectations even in automated settings. Agnese Palazzi argues that the automation of public authority does not diminish the importance of lawyers but intensifies it. Through the concepts of micro-agency and semantic capital, she shows how lawyers providing legal representation in proceedings involving automated systems play a crucial and irreplaceable role in safeguarding effective judicial protection. Janko Munjić traces the impact of automated tools across the criminal justice process, from predictive policing through risk scoring to courtroom evidence, arguing that the cumulative effect is a slow normalization of outputs that feel objective but compound existing inequalities. He closes with a ‘minimum humanity package’: a baseline of conditions that must hold if criminal justice is to remain recognizably human. María Manuela Márquez Velásquez and José Restom turn to the International Criminal Court as a distinctive institutional setting in which AI is already being deployed for evidence processing, data analysis, and case management. They analyse the specific risks this poses, including bias, loss of control, and procedural harm, and propose governance principles for responsible AI use in international criminal proceedings.
Dignity, Surveillance, and the Human Subject
The fourth theme steps back from institutional mechanics to ask what algorithmic governance does to the human subject: to dignity, autonomy, and the conditions of a fully human life. Behind every system, audit mechanism, and legal remedy lies a person. This theme asks what kind of person the automated state makes room for.
In the lecture series, Sofia Ranchordás (Tilburg University/Luiss Guido Carli) showed how efficiency-driven datafication has displaced solidarity, empathy, and individual vulnerability from the logic of public administration. Christine Moser (VU Amsterdam) argued that moving beyond polarized narratives of salvation and destruction opens space for more grounded, constructive engagement with algorithmic systems.
Three contributions in this Symposium pursue these questions. Ali Mert Gürkan maps the ethical assumptions that underlie contemporary regulatory and industry responses to algorithmic governance: assumptions of procedural sufficiency, statistical objectivity, and predictive utilitarianism. He argues that these assumptions reshape the relationship between institutions and individuals in ways that fundamentally threaten human dignity and autonomy, and that understanding them is a precondition for more adequate responses. Antoni Napieralski uses empathy as a method of legal interpretation to analyse the automated surveillance regime applied to third-country nationals at the EU’s external borders. He shows how citizenship operates as a condition of surveillance intensity, creating a two-tier structure of legal protection in which non-citizens are subjected to broader retention, wider access, and more durable traceability than EU citizens, despite the formally universal scope of the right to data protection. Öznur Uğuz examines how emotional AI surveillance, systems that detect, infer, and interact with human emotions, threatens human dignity by interfering with autonomous decision-making and reshaping behaviour through coercion or manipulation. Drawing on Kantian ethics, she argues that emotional AI surveillance objectifies individuals as mere means to others’ ends, constituting a direct violation of an autonomy-centred understanding of human dignity.
The Questions that Remain, on Both Sides of the Machine
The contributions gathered in this Symposium illuminate a profound transformation: one in which automation is quietly reshaping not only how the state operates, but what it means to be human on both sides of public administration. Systems built around efficiency, speed, and scalability increasingly treat humanity as a residual concern rather than a central design principle. Administrators find their judgment displaced to the margins, while those subject to automated decisions experience growing alienation from the very institutions meant to serve them.
As the Symposium essays explore in depth, the legal, institutional, and democratic tools we rely on were forged for human-to-human encounters. Whether these frameworks can be meaningfully adapted to the automated state or whether we must build new ones, grounded in a clearer understanding of human experience, remains the central challenge. The question is no longer just technical, but also existential: what kind of humans can we be within institutions that increasingly act upon people, without people? In a concluding post, we return to these questions, tracing across the full range of contributions where the human keeps getting displaced, what that displacement costs, and what, if anything, remains to be built.
Daria Morozova
Daria Morozova is Assistant Professor at the Department of Business Studies at Leiden Law School. Her research explores what it means to be human at work, examining how people perceive their capacities in the artificially intelligent world, what makes people successful in changing environments, and how the ways we speak about the world shape our experiences.
Melanie Fink
Melanie Fink is Associate Professor of European Law at Leiden University. Her research explores the intersection of digital public administration and EU law from an individual rights perspective.
