Governing by Algorithms: Eduardo Albrecht’s Diagnosis of Institutions in the Age of AI
Key Notes
AI systems embedded in public administration are increasingly exercising decision-making authority at a speed and scale that existing democratic institutions were not designed to govern at.
The growing fusion of public institutions and private technology firms has diffused political power and weakened traditional mechanisms of accountability and citizen influence.
Proposals for new institutional forms, such as a Third House and digitally mediated civic participation, emerge from the claim that democratic legitimacy cannot be sustained without adapting institutional design to algorithmic governance.
Introduction
The central concern is whether contemporary democratic institutions are structurally capable of governing computational systems that operate at unprecedented speed and scale.
Across governments, algorithmic systems are now embedded in routine administrative functions, shaping decisions related to welfare eligibility, policing, border control, public communication, and national security. Much of the public debate surrounding this shift remains focused on familiar concerns such as algorithmic bias, transparency, and regulatory oversight. While these issues are important, they do not fully capture the depth of the transformation underway in governance and democracy.
In his recent book Political Automation: An Introduction to AI in Government and Its Impact on Citizens (Oxford University Press, 2025), Professor Eduardo Albrecht explores the tension between technology and government. Rather than treating artificial intelligence as a tool to be regulated within existing democratic frameworks, he approaches it as a structural force that is quietly reconfiguring political power itself. In conversation, Albrecht consistently resists ideological positioning: he does not present himself as a technologist advocating new solutions, nor as a critic seeking to halt technological adoption. Instead, he frames his work as diagnostic. The central concern is whether contemporary democratic institutions are structurally capable of governing computational systems that operate at unprecedented speed and scale.
Disclaimer: this essay offers an account of Albrecht’s thinking based on an extended interview around his latest book, Political Automation - Eduardo Albrecht - Oxford University Press. It does not seek to evaluate the desirability of his proposals or to resolve their feasibility. Its purpose is to clarify how he conceptualizes the political implications of artificial intelligence, why he believes existing institutions are under strain, and what unresolved questions his framework raises for democratic governance.
A Diagnostic Framework: Speed and Ubiquity
Albrecht’s research and conclusions do not emerge from ideological commitments but from a diagnostic assessment of institutional performance. Before prescribing institutional reform, he argues, it is necessary to identify what is not functioning within existing democratic structures.
From this diagnostic perspective, democratic institutions encounter structural limits in governing artificial intelligence for two primary reasons. The first is speed. Representative governance operates through extended cycles of deliberation, legislation, and implementation. Public debate, committee review, and electoral accountability remain essential to democratic legitimacy, yet they impose temporal constraints that algorithmic systems do not share. By contrast, AI deployment in public administration advances at a pace that exceeds institutional decision-making cycles by orders of magnitude. Algorithmic systems are deployed, updated, and scaled far more rapidly than legislative or oversight processes can respond.
The second constraint is ubiquity. Artificial intelligence in governance does not appear as a discrete or bounded set of applications that can be catalogued and regulated individually. Instead, these systems proliferate across agencies, jurisdictions, and domains in an exponential and decentralized manner. Within this framework, existing institutions resemble a net too small to capture the volume and dispersion of political machines operating throughout society. In this context, political machines refers to policy decisions made partly or entirely with the support of algorithmic systems. Even well-intentioned oversight mechanisms struggle to identify, let alone govern, the full scope of algorithmic decision-making that now shapes citizen outcomes.
Taken together, speed and ubiquity form the core of Albrecht’s diagnosis. The problem, as he frames it, is not simply regulatory lag or technical opacity. It is a structural mismatch between institutional design and technological reality.
Political Machines and the Locus of Power
A central feature of Albrecht’s framework is his insistence that artificial intelligence should be understood as a locus of political power. This reframing distinguishes his work from debates that focus primarily on ethics or compliance. When decision making is delegated to algorithmic systems, he argues, computational infrastructures that operate beyond the direct visibility of citizens subsume political authority.
This authority is not centralized in a single institution. Instead, it is diffused across bureaucracies, software systems, and public-private arrangements. As a result, accountability becomes difficult to locate. Citizens often experience the effects of algorithmic decisions without knowing where those decisions originate or who is responsible for them. In Albrecht’s view, this diffusion contributes to broader patterns of political anxiety and declining trust in democratic institutions.
He does not attribute the current crisis of legitimacy solely to artificial intelligence. Polarization, populism, and digital media all play roles. However, he suggests that the quiet relocation of decision making into opaque computational systems exacerbates these dynamics by weakening traditional mechanisms of representation and oversight.
Public-Private Fusion and Corporatist Dynamics
The diffusion of power is further complicated by the growing interdependence between governments and private technology firms. Albrecht emphasizes that the regulatory response to artificial intelligence depends in part on who is using it. Governmental use traditionally triggers public law constraints, while private use is governed by market regulation and liability frameworks. In practice, however, this distinction is increasingly blurred.
Many algorithmic systems deployed by governments are commissioned by public institutions but designed, built, and maintained by private companies. These public-private partnerships make it difficult to determine where authority resides and which accountability mechanisms apply. To conceptualize this dynamic, Albrecht draws on a corporatist model in which state and corporate actors merge into a powerful hybrid entity.
He is careful to note that historical analogies should not be overstated. His reference to corporatist systems in twentieth century Europe is not an assertion of equivalence. Rather, it serves to illustrate a recurring structural equation. When state capacity and corporate infrastructure combine, power increases, often at the expense of individual citizens. In this configuration, the capacity of citizens to influence political outcomes diminishes, not necessarily through overt repression, but through structural marginalization.
The Third House as an Institutional Hypothesis
It is within this diagnostic context that Albrecht introduces the idea of a “Third House.” He does not present it as an imminent policy proposal or a fully specified institutional design. Instead, he frames it as a conceptual response to the structural mismatch he identifies.
“The Third House” is envisioned as an institutional layer capable of operating at the speed and scale of political machines, characterized as a virtual chamber that supplements existing bodies such as the Senate and the House. Its purpose would be to restore a degree of citizen agency by mediating between algorithmic systems and democratic oversight. Albrecht emphasizes that this idea emerges from diagnosis rather than ideology. He explicitly welcomes disagreement and alternative proposals.
Crucially, he rejects the notion that such an institution could be imposed from above. In his account, political institutions historically emerge from bottom up processes, often following periods of crisis. He anticipates that any movement toward something resembling a Third House would begin with small scale experiments, particularly at the local and municipal level, before potentially gelling into actual institutions.
In this context, Albrecht argues that the expansion of an AI-driven economy further intensifies the need for a Third House, noting that existing democratic frameworks were not designed to support economic systems that increasingly depend on automated decision-making. Albrecht encapsulates this view in a single formulation: “We will not be able to have this new AI-driven economy without some form of a new institution that supports the existence of this new AI-driven economy.” The statement reflects his belief that institutional adaptation is not optional once automated systems become foundational to public and private institutions.
Digital Citizens and the Limits of Representation
The most contested element of Albrecht’s framework concerns digital citizens, or citizen avatars. These are AI generated citizen "doubles" that represent the real person in AI-based "deliberations" that occur in the virtual chamber, or Third House.
He acknowledges that this proposal generates strong resistance, as it touches on deeply held intuitions about personal agency and political voice. The idea that individuals might delegate aspects of civic participation to algorithmic representatives raises immediate concerns about trust, control, and identity.
Albrecht insists that his proposal does not stem from technological enthusiasm. Rather, he frames it as a pragmatic response to the limits of human participation in an environment dominated by machine speed and scale. In his account, citizen avatars would function as personal emissaries, representing an individual’s articulated preferences within institutional processes that humans alone cannot feasibly monitor.
A key condition, in his view, is tethered control. He uses the metaphor of autonomous vehicles to illustrate this relationship. While the system may operate independently, the human remains responsible for setting direction and defining parameters. Even so, Albrecht acknowledges that the technical, social, and cultural challenges of such a system are substantial. He does not claim that these issues have been resolved, only that they cannot be avoided if citizen agency is to remain meaningful.
Rights as Preconditions: Data Access and Freedom of Thought
Albrecht situates his institutional thinking within a broader rights-based framework. He identifies two legal developments as prerequisites for any viable response to algorithmic governance. The first is comprehensive access to personal data. He argues that citizens should be able to know what data governments and public private partnerships hold about them, how they are categorized, and how those categories affect access to resources and rights.
The second is the protection of freedom of thought. As algorithmic systems increasingly influence perception, classification, and decision making, Albrecht contends that cognitive autonomy becomes a central political concern. The erosion of privacy, particularly in intimate and cognitive domains, contributes to widespread mistrust and anxiety. In this sense, rights are not merely safeguards but structural conditions for legitimacy.
The Institutional Questions That Remain
Albrecht’s framework raises difficult and unresolved questions. Can democratic institutions evolve quickly enough to govern computational systems without sacrificing legitimacy? Can representation be mediated through technology without alienating citizens from their own political voice? Can accountability be restored in a landscape where power is diffused across invisible infrastructures?
This essay does not attempt to answer these questions. Its purpose has been to render a particular way of thinking legible. Whether one agrees with Albrecht’s conclusions or not, his work reflects a growing recognition that artificial intelligence is not merely a technical challenge: it is an institutional one. Understanding that distinction may be a necessary first step in navigating the political transformations now underway.
