BEYOND REPAIR
Rethinking Democratic Legitimacy Under Epistemic Constraint — Working Paper No. 4
THE LEGITIMACY PROJECT — Working Paper No. 4
BEYOND REPAIR
Rethinking Democratic Legitimacy Under Epistemic Constraint
The erosion of epistemic conditions does not eliminate the need for governance. It eliminates the comfort of inherited assumptions about how governance is justified.
By The Legitimacy Project | April 2026 | Working paper
I. FROM INSTITUTIONAL DESIGN TO THEORETICAL REVISION
The first three working papers in this series established a sequence. Democratic legitimacy presupposes epistemic conditions. Those conditions are failing under systematic structural pressure. The repair project, which attempts to restore them from within the democratic framework, faces constraints that are deep enough to cast doubt on its sufficiency at democratic scale. The present paper asks what follows from that sequence if the doubt is warranted.
The repair project operates within democratic theory. When it confronts the limits identified in working paper 3, it tends to respond by refining the repair strategy rather than questioning the framework. That response is reasonable if the gap between what democratic legitimacy requires and what the current information environment can produce is bridgeable by institutional means. The analysis in working paper 3 suggests it may not be. If that assessment holds, the problem shifts from institutional design to theoretical revision, and the question becomes how to think about legitimacy under conditions where its traditional foundations do not reliably obtain.
This paper maps the space of possible responses without advocating for any of them. The appropriate analogy is to a systems theorist examining design options under a given set of constraints, rather than to a political theorist defending a preferred model. Each response involves tradeoffs that are real and significant, and the paper’s aim is to make those tradeoffs explicit. Identifying the available options and their costs is a precondition for any serious subsequent evaluation, and that evaluation has not yet been adequately undertaken in the existing literature.
I proceed as follows. Section 2 specifies the constraints that any viable response must satisfy. Sections 3 through 7 examine five revision paths in order of how much they preserve the participatory basis of democratic legitimacy. Section 8 identifies the tradeoffs that cut across all five paths. Section 9 develops the deeper theoretical problem that revision raises. Section 10 states the position of this project in relation to that problem.
II. WHAT ANY VIABLE SYSTEM MUST ADDRESS
Before examining revision paths, it is necessary to specify what any viable governance system must accomplish, independent of its particular structure. This is a question about functional requirements, specifically what any system that aspires to govern a modern political community must meet to remain stable and justifiable over time.
Any viable system must provide mechanisms for collective decision-making on questions that affect the population as a whole, since the coordination problems that governance exists to solve do not disappear under epistemic constraint. It must generate at least a working basis of legitimacy, meaning a reason, even if attenuated or contested, why those subject to decisions accept their binding force rather than simply resist. It must maintain some relationship to epistemic adequacy, in the sense that decisions must be connected, however imperfectly, to accurate information about the conditions they are meant to address. And it must be stable enough to persist over time without constant crisis and renegotiation.
These requirements pull against each other in ways that the revision paths below make visible. A system that maximizes epistemic adequacy may sacrifice legitimacy. One that maximizes legitimacy under non-convergent epistemic conditions may sacrifice epistemic performance. The task is to understand the tradeoff structure, not to dissolve it.
III. REVISION PATH 1: EPISTEMICALLY CONSTRAINED DEMOCRACY
The revision path most continuous with existing democratic theory introduces epistemic constraints on democratic participation without abandoning the participatory framework. Citizens retain formal equality before the law and in electoral processes, but deliberative mechanisms are structured to filter out or weight down the influence of epistemically degraded inputs. The clearest contemporary expressions of this path are deliberative mini-publics, randomly selected citizen assemblies that deliberate on constrained agendas under conditions designed to favor informed discussion over reactive preference expression.1
The appeal of this path is that it attempts to restore the epistemic conditions democratic legitimacy requires by restructuring the deliberative process rather than by altering the information environment. It accepts that the broader epistemic environment cannot be repaired and attempts to create insulated deliberative spaces that are less subject to the selection pressures described in working paper 2. Within those spaces, the conditions for genuine deliberation, including shared information, expert input, and sustained reasoning under low cognitive load, are manufactured by design.
The tradeoff is between epistemic performance and equality of participation. Deliberative mini-publics improve the quality of deliberation by limiting participation to a randomly selected subset of the population and by structuring their information environment in ways that the broader population does not experience. The legitimacy of their outputs depends on the representativeness of the selection and on the acceptance of their authority by those not included. Both of these conditions are fragile in exactly the conditions of trust fragmentation that make repair difficult. A population that distrusts expert institutions has limited grounds for trusting the outputs of a deliberative process designed and administered by those institutions.2
IV. REVISION PATH 2: INFORMATION-GATED PARTICIPATION
A more direct response to the epistemic problem ties participation to demonstrated informational competence. On this path, some form of qualification, whether verified knowledge of the issues at stake or assessed reasoning ability, mediates access to consequential decision-making. The idea has a long history in political philosophy, from Mill’s plural voting scheme to contemporary proposals for epistocracy, and it is the revision path that most directly targets the mechanism of epistemic failure.3
If the problem is that democratic decision-making aggregates preferences formed under epistemically distorted conditions, a system that conditions participation on epistemic competence addresses the problem at its source. Decisions would be made by citizens whose beliefs are at least minimally reliable, and the aggregation problem that working paper 1 identified, that the vote may no longer track what legitimizes it, would be at least partially dissolved.
The design of competence criteria is itself a political act, and whoever controls that design controls which citizens participate. The history of literacy tests and similar qualification mechanisms demonstrates that epistemic criteria are reliably available to be shaped by existing power structures to exclude the populations those structures prefer to exclude.4
Even a well-designed competence criterion faces the further problem that competence, in a non-convergent epistemic system, is itself contested. If citizens disagree about what counts as reliable evidence, they will disagree about what counts as knowledge, and the authority to adjudicate that disagreement must come from somewhere. That authority shift problem, the question of who legitimately decides what competence requires, is the deepest difficulty this path faces and the one its proponents have been least willing to confront directly.
V. REVISION PATH 3: TECHNOCRATIC AUGMENTATION
Technocratic augmentation increases the role of experts and specialized institutions in political decision-making, shifting authority along a spectrum from advisory to influential to determinative. In its weaker forms, this path is already embedded in democratic governance, regulatory agencies, central banks, and judicial bodies all exercise forms of insulated authority over consequential decisions. The revision path under consideration extends that insulation more broadly, in response to the argument that mass participation under epistemically degraded conditions produces systematically worse outcomes than decisions made by smaller groups of appropriately qualified actors.5
If citizen belief formation is subject to the structural distortions described in working papers 1 through 3, and if those distortions systematically bias democratic outputs away from accurate assessments of policy problems, then there are outcome-based grounds for increasing the epistemic quality of decision-making by reducing the role of epistemically compromised inputs. The argument requires only that expert institutions in domains where expertise is well-defined and verifiable produce more reliable outputs than mass deliberation under non-convergent epistemic conditions.
The tradeoff is between epistemic performance and legitimacy, and it is more severe on this path than on the previous two. Technocratic authority derives its legitimacy from a combination of demonstrated competence and democratic delegation. The public accepts expert authority because it has been granted by democratic processes and remains accountable to them. Both sources of legitimacy are under strain in conditions of trust fragmentation. Expert competence is contested by populations whose information environments have systematically cultivated that contestation. Democratic delegation is weakened when the delegating processes are themselves epistemically degraded.6
Technocratic augmentation may improve decision quality at the cost of the legitimacy conditions that make those decisions politically sustainable.
VI. REVISION PATH 4: MANAGED INFORMATION ENVIRONMENTS
Managed information environments attempt to address the epistemic problem indirectly, by restructuring the conditions under which citizens form beliefs rather than by altering the structure of participation itself. On this path, legitimate authority actively shapes the information environment through curation, filtering, or restriction, with the aim of reducing the epistemic selection pressure that working paper 2 identified as the primary mechanism of deterioration. The goal is to restore the epistemic conditions democratic legitimacy requires rather than to work around their absence.
This path is distinct from the repair project examined in working paper 3. The repair project attempts to restore epistemic conditions through interventions that are themselves subject to democratic authorization and constrained by democratic norms, including norms of free expression. Managed information environments involve a more direct exercise of authority over the epistemic environment, one that may require setting aside those norms to achieve its aims. The distinction matters because the self-undermining problem identified in working paper 3 is most acute on this path. An authority that controls the information environment controls the conditions under which its own legitimacy is assessed. That circularity is potentially manageable in environments of high trust. In conditions of trust fragmentation, it is politically explosive.7
The appeal of this path is that it targets the correct level of the problem. If epistemic conditions are structural rather than individual, then structural interventions are appropriate and individual-level approaches, including deliberative mini-publics and competence criteria, will be insufficient. The cost is that structural intervention in the information environment concentrates the authority to define epistemic quality in whoever administers the intervention, and the history of such authority suggests that it is reliably subject to capture by the political interests of the administering body. The path toward epistemic restoration and the path toward epistemic control are, from the outside, difficult to distinguish.
VII. REVISION PATH 5: DECENTRALIZED GOVERNANCE
The four revision paths examined so far attempt, by different means, to maintain or restore the epistemic conditions that legitimate governance requires at the scale of modern democratic states. Decentralized governance takes a different approach. Instead of attempting to produce epistemic convergence across large and heterogeneous populations, it accepts the reality of non-convergent epistemic systems and responds by reducing the scale at which governance must operate.
If citizens inhabiting different information ecosystems share insufficient epistemic common ground for legitimate collective decision-making at national scale, the response may be to devolve consequential decisions to smaller political units where epistemic communities are more homogeneous and shared informational baselines more achievable. Within such units, the preconditions for democratic legitimacy are at least partially recoverable. The coordination and aggregation problems that make democratic legitimacy difficult at scale are reduced when the population is smaller and more epistemically coherent.8
The costs of this path are concentrated in what it gives up. Many of the governance problems that modern states must address, like climate change, pandemic response, financial stability, and security, are not tractable at the scale of epistemically homogeneous local communities. Decentralization that is genuinely responsive to the epistemic problem would require accepting a significant reduction in the scope of collective decision-making. The problems that require coordination across epistemically divergent populations would either go unaddressed or be addressed by mechanisms that are not themselves democratically legitimate at the required scale.
VIII. WHAT EVERY PATH COSTS
The five revision paths involve tradeoffs that are structural, meaning that movement toward one value requires accepting costs on another. Making those tradeoffs explicit is the precondition for evaluating the paths honestly.
The central tradeoff runs between epistemic quality and equality of participation. Every path that improves the epistemic quality of collective decision-making does so by reducing the influence of epistemically compromised inputs. Reducing that influence means reducing the equal standing of the citizens providing those inputs. The democratic commitment to political equality and the epistemic commitment to reliable belief formation are, under conditions of degraded epistemic environments, in genuine tension. There is no revision path that resolves that tension rather than accepting a particular position within it.
A second tradeoff runs between legitimacy and performance. Governance systems derive their legitimacy partly from outcomes and partly from process. Epistemic constraint on the process tends to improve performance at the cost of process legitimacy. Epistemic openness tends to preserve process legitimacy at the cost of performance. The conditions under which a governance system can have both are precisely the conditions that working papers 1 through 3 have established are currently absent.
A third tradeoff runs between stability and justificatory coherence. A governance system that maintains democratic forms while quietly concentrating epistemic authority in insulated institutions may be more stable than one that openly acknowledges the tradeoffs it has made. A system that is stable only because its operating logic is not publicly acknowledged is, however, justificatorily incoherent. The gap between the democratic self-description and the technocratic reality is itself a source of epistemic distortion, producing citizens whose beliefs about their own governance system are systematically inaccurate, a condition generated by the governance system itself.
IX. WHAT REPLACES PARTICIPATION AS THE SOURCE OF LEGITIMACY
The revision paths and their tradeoffs converge on a theoretical problem that none of them resolves. Democratic legitimacy, in its canonical formulations, derives from the equal participation of citizens in collective self-governance. Participation under conditions of shared epistemic access generates decisions that citizens can recognize as, in some sense, their own, even when they disagree with particular outcomes. It is this participatory basis that gives democratic legitimacy its distinctive character and that distinguishes it from legitimacy grounded in outcomes alone, or in the historical authority of tradition.9
Every revision path examined above either reduces participation, conditions it, or restructures the informational environment in which it occurs. In each case, the participatory basis of legitimacy is attenuated. The question that follows is whether any of the remaining sources of legitimacy are adequate substitutes. Outcomes are insufficient on their own, because a governance system whose legitimacy depends entirely on producing good outcomes has no basis for legitimacy during the periods when its outputs are contested or poor. Expertise is insufficient on its own, because the authority problem means that the designation of expertise requires a legitimating basis that expertise itself cannot provide.
The theory of legitimacy adequate to the current condition has not yet been developed. The revision paths map the available design space, but they do not solve the theoretical problem of what justifies governance when its traditional foundations are absent. That problem is the most consequential open question in contemporary political theory, and it is one the field has been slow to engage with the directness it requires, partly because engaging with it directly requires acknowledging that the epistemic conditions democratic theory assumes are not reliably present.
X. AN INVESTIGATION WITHOUT A CONCLUSION
The Legitimacy Project does not begin from a commitment to any of the revision paths examined above. Its aim across these four working papers has been to establish the problem with the precision and seriousness it deserves, as a precondition for evaluating the available responses. The sequence of preconditions, mechanism of breakdown, limits of repair, space of revision is an investigative structure aimed at clearing the ground rather than arriving at a predetermined destination.
The evaluation of revision paths requires criteria that this project treats as empirical questions rather than as normative commitments: feasibility under actual technological and institutional conditions, stability over time given the incentive structures that actually operate, and justificatory coherence in the sense that the operating logic of the system is at least in principle publicly defensible to those subject to it. These are demanding criteria, and none of the revision paths examined above satisfies all three with confidence. That is the finding, and it is an uncomfortable one.
The discomfort is appropriate, as a political theory that presupposes epistemic conditions that cannot be reliably restored must eventually confront that fact. The alternatives to confronting it are to ignore the problem, which produces governance systems that are less legitimate than they appear, or to acknowledge it only partially, which produces governance systems whose self-descriptions are systematically misleading. Neither alternative is satisfactory, and the cost of each is borne primarily by the populations subject to governance rather than by the theorists describing it.
What replaces the inherited assumptions remains an open question. The Legitimacy Project’s contribution is to establish that it is no longer a hypothetical one.
REFERENCES
Achen, Christopher H. and Larry M. Bartels, 2016, Democracy for Realists, Princeton: Princeton University Press.
Anderson, Elizabeth, 2006, “The Epistemology of Democracy,” Episteme, 3(1-2): 8-22.
Balkin, Jack M., 2018, “Free Speech in the Algorithmic Society,” UC Davis Law Review, 51(3): 1149-1210.
Brennan, Jason, 2016, Against Democracy, Princeton: Princeton University Press.
Caplan, Bryan, 2007, The Myth of the Rational Voter, Princeton: Princeton University Press.
Christiano, Thomas, 2008, The Constitution of Equality: Democratic Authority and Its Limits, Oxford: Oxford University Press.
Cohen, Joshua, 1989, “Deliberation and Democratic Legitimacy,” in Alan Hamlin and Philip Pettit (eds.), The Good Polity, Oxford: Basil Blackwell, 17-34.
Estlund, David, 2008, Democratic Authority: A Philosophical Framework, Princeton: Princeton University Press.
Fishkin, James S., 2018, Democracy When the People Are Thinking, Oxford: Oxford University Press.
Hayek, Friedrich A., 1945, “The Use of Knowledge in Society,” American Economic Review, 35(4): 519-530.
Kolodny, Niko, 2014a, “Rule Over None I: What Justifies Democracy?” Philosophy and Public Affairs, 42(3): 195-229.
Kolodny, Niko, 2014b, “Rule Over None II: Social Equality and the Justification of Democracy,” Philosophy and Public Affairs, 42(4): 287-336.
Lafont, Christina, 2019, Democracy Without Shortcuts, Oxford: Oxford University Press.
Landemore, Helene, 2013, Democratic Reason: Politics, Collective Intelligence, and the Rule of the Many, Princeton: Princeton University Press.
Mill, John Stuart, 1861 [1991], Considerations on Representative Government, Buffalo: Prometheus Books.
Peter, Fabienne, 2009, Democratic Legitimacy, New York: Routledge.
Rousseau, Jean-Jacques, 1762, The Social Contract, Charles Frankel (trans.), New York: Hafner, 1947.
Zittrain, Jonathan, 2019, “The Problem of Infodemic: How Dominant Platforms Can Serve the Public Interest,” in Disinformation, ‘Fake News’ and Influence Campaigns, Matthew Bunker (ed.), London: Routledge.
For deliberative mini-publics as a reform mechanism, see Lafont (2019) and Fishkin (2018) on deliberative polling. The most extensive empirical record comes from citizens’ assemblies on constitutional questions in Ireland and Iceland, which provide evidence on both the epistemic performance and the legitimacy conditions of this format.
Lafont (2019) develops the most sustained critique of mini-publics from within deliberative democratic theory, arguing that they risk substituting the deliberation of a few for the ongoing public reasoning of the many. The trust fragmentation problem compounds this concern: outputs from mini-publics are no more insulated from contested authority than outputs from expert institutions.
Mill’s plural voting proposal is developed in Considerations on Representative Government (1861). For contemporary epistocracy proposals, see Brennan (2016), who provides the most systematic recent defense. For the strongest systematic response, see Landemore (2013) on the epistemic advantages of inclusive participation.
Anderson (2006) documents the history of epistemic criteria for political participation and their use as exclusionary mechanisms. The argument that competence criteria are structurally available to be captured by existing power relations is developed in Estlund (2008), who uses it as a constraint on what he calls “epistocracy.”
For the empirical case for insulated expert authority, see Achen and Bartels (2016) on the gap between democratic theory’s assumptions about citizen competence and the empirical record. For the normative case for technocratic authority under epistemic constraint, see Caplan (2007), whose argument for the rationality of rational ignorance is relevant even if his conclusions are contested.
Christiano (2008) develops the relationship between democratic legitimacy and epistemic conditions in ways that bear on this tradeoff. The specific argument that technocratic authority faces a compounded legitimacy problem under trust fragmentation draws on the analysis in working paper 3.
Balkin (2018) and Zittrain (2019) develop the concept of epistemic authority over information environments and the legitimacy conditions under which it can be exercised. The distinction between repair-oriented and control-oriented information management is implicit in their analyses.
The decentralization response to epistemic fragmentation has affinities with Hayek’s (1945) argument for decentralized decision-making as a response to the knowledge problem, though the epistemic problem addressed here is different in kind: the issue is the reliability of belief formation, not the distribution of local knowledge.
The participatory basis of democratic legitimacy is developed most fully in Rousseau (1762), Cohen (1989), and Christiano (2008). For the specific argument that participation generates a distinctive form of legitimacy not reducible to outcomes or expertise, see Kolodny (2014a, 2014b).