Philosophers and political commentators sometimes suggest that randomly chosen citizens may be able to replace democratically elected legislators in whole or in part (Abizadeh 2020; Guerrero 2014, 2021a, 2021b; Landemore 2020; Owen and Smith 2018; Vandamme and Verret-Hamelin 2017). Arguments are commonly a blend of moral, political and epistemic considerations, reflecting the fact that elected representatives are often themselves inexpert on the matters on which they have to decide, and that their selection may be the consequence of factors that are arbitrary epistemically and morally, although overdetermined politically once one considers the distribution of power and resources in society.

However, it is unclear that moral, political and epistemic considerations single out lotteries as the privileged way to select political representatives. Even if lotteries were epistemically superior to either appointment or elections, there might be compelling moral and political reasons to prefer the latter when selecting people for legislative office, and to favour appointment over both elections and lotteries when selecting legislative advisors (Landa and Pevnick 2021; Umbers 2021). Efforts to create more directly participative forms of local government may be democratically justified even if they are epistemically imperfect or less efficient than other forms of local government, given the importance of ideas of self-government to democratic ideas and ideals (Dugrand 2020; Stone 2021a). So why suppose that democrats should favour lotteries over elections on epistemic grounds?

To answer that question, let us look at Alexander Guerrero’s epistemic case for ‘lottocracy’, or government by randomly selected citizen assemblies (Guerrero 2014, 2021b, 2021a). Efforts to develop epistemic arguments for democracy tend to treat elections as a defining mechanism of democracy, and to suppose that the personal, civil and political rights required to distinguish democratic from undemocratic elections are essential to the epistemic virtues of democracy (Misak and Talisse 2021; Talisse 2013; Lever and Chin 2019; Gerber 2021; Lever 2021). By contrast, Guerrero is not particularly interested in epistemic comparisons between democratic and undemocratic government and, instead, is chiefly concerned with dethroning the centrality of elections to contemporary democratic thought and practice. He believes that citizens can be sources of expertise—moral as well as technical and scientific—and that randomly selected citizens, invested with legislative power and authority, are an attractive and useful way to tap into that expertise.

As we will see, Guerrero fails to show that randomly selected legislatures would be an improvement over elected ones epistemically—or morally and politically for that matter. Moreover, the right of democratic citizens to deliberate together, and their claims to adequate information to do so, do not depend on showing that they are especially wise individually or collectively. No such claims are necessary on democratic grounds. Hence, I argue, while Guerrero has not shown that we should replace elected with randomly selected citizens, advisory sortition assemblies might play a variety of useful roles in democratic politics.

For the purposes of this paper ‘citizen assemblies’ refers to a deliberative body composed of randomly selected citizens selected to debate a matter of public interest. I therefore include in this description small ‘minipublics’, larger citizen assemblies that are meant to be purely deliberative and those, such as Guerrero advocates, which are a combination of single-issue legislative bodies or SILLS.Footnote 1 I agree with Keith Dowding in this issue (pp…) that expertise is not best thought of in terms of the probability of getting the right answer to a problem, although Guerrero tends to adopt the ‘right answer’ approach, as do Elster and Holst (pp….). Insight and creativity, nuance and clarity, facility and effortlessness are also connected to our ideas of expertise, in moral as in scientific matters. However, nothing in this paper depends on adopting Dowding’s particular conception of expertise, and exemplars of moral wisdom may be harder to detect than he implies both because opportunities to demonstrate it may be scarce, and because real expertise is often unshowy.

Guerrero and the Epistemic Case Against Elections

According to Alexander Guerrero, the use of elected representation is typically defended on epistemic and agential grounds and ‘embod[ies] a kind of compromise’ between these values.Footnote 2 For representative government to be justified agentially or epistemically, he believes, you need ‘meaningful accountability’ which, in turn, presupposes practices of informed monitoring and evaluation at every point in the representative chain (p. 161).Footnote 3 Such meaningful accountability is undermined by citizens’ pervasive and deep ignorance about issues of public policy, the workings of government and their political circumstances (pp. 159–160). As a result, officials are readily ‘captured’ by the groups that they are supposed to regulate (pp. 160–161), and citizens regularly elect representatives whose lives are quite different from their own and likely to generate conflicts of interest between governed and governors (pp. 161–162, 167–168). Democratic outcomes therefore tend to be epistemically poor; and these poor outcomes are exacerbated over time by the epistemically perverse incentives created by elections. Chief amongst those perverse incentives are short-termism (pp. 164–165), dubious practices of blame-shifting and credit-claiming as well as the appeal to, and manipulation of, citizens’ emotions (pp. 165–166) in the interests of poorly thought out, counterproductive and, even, dangerous partisan agendas. In short, Guerrero claims, ‘Voter ignorance undermines meaningful electoral accountability. An absence of meaningful electoral accountability results in capture. And capture results in what might well be described as epistemic disaster’ (p. 163).

By contrast, Guerrero believes, a lottocratic system of government—particularly one organised based on multiple single-issue legislatures or SILLS, as he calls them—enables us to avoid the epistemic defects of elections without abandoning democracy ‘for epistocracy’ (p. 169).Footnote 4 Random selection enables a diverse group of citizens to participate in collective deliberation and decisions, based on learning from pre-selected experts—though Guerrero says nothing about how these experts are to be selected, by whom and what weight assembly members are meant to put on expert opinion, as opposed to their own or each others’. Hence, according to Guerrero, citizens do not need to be particularly wise or virtuous to be capable legislators. Because citizens are randomly chosen and will be changed regularly via rotation, they cannot be captured by special interests, and will tend to be largely impartial and fair-minded when considering public issues (p. 175). Finally, because citizens will be randomly selected, they will lack the hyper-partisanship that bedevils party-based electoral systems. They will also demonstrate epistemic humility (p. 175), because they cannot be said to have deserved their place in the legislature. In short, it seems that lottocratic political institutions foster equality in virtue and direct that equal virtue to politically useful purposes, while elections render citizens and representatives morally, politically and epistemically vicious. Democracy, then, requires us to replace elections with randomly selected legislative assemblies on epistemic grounds that also reflect the moral and political virtues of lotteries, as compared to elections. Hence Guerrero’s critique of electoral democracy is meant to be compatible with, and to embody, democratic moral and political principles in ways that would not be true of epistocracy.Footnote 5

The Limitations of the Epistemic Case for Lotteries

Interesting though it is, Guerrero’s epistemic case for replacing elections by lot is not persuasive. Normatively, a lottocratic government seems ill-suited to democratic ideals of co-governance or co-rulership. The fact that everyone will have a formally equal, though very small, chance of being randomly selected for power scarcely answers to the democratic ideal that citizens should be able to see themselves as the authors, not just the addressees of law (Beerbohm 2012). For all their defects, elections enable people collectively to decide who should form part of their legislature; create opportunities for citizens to influence the selection of candidates for office indirectly via their votes and, where primaries are used to select the candidates for office, citizens get to influence the choice of candidates more directly as well (Stone 2021a). Above all, elections guarantee citizens a role in politics as voters, without committing to a career of public service or political engagement. Proportional representation, as distinct from First Past the Post, magnifies that role, by ensuring that nearly all citizens will have someone in the legislature who they personally support. The right to stand as a candidate for legislative office, in turn, means that citizens are entitled to seek greater public responsibilities at a time of their choosing, or in response to requests by others. By contrast, lottocracy is intentionally indifferent to citizen interests in shaping the political agenda collaboratively, or in deciding which candidates for their trust and confidence are entitled to it. In these ways lottocracy seems an unsatisfactory expression of people’s interests in political agency—whatever one’s view of elections (Lafont 2020; Ceva and Ottonelli 2021; Umbers 2021; Girard 2019; Destri and Lever 2023; Stone 2016). Thus Guerrero’s confidence that randomly selected representatives will be humble seems rather a testimony to the limited agency that lottocracy grants citizens than a confident prediction of how members of a sortition assembly will behave once they grapple with the demands of office.Footnote 6

More seriously from Guerrero’s own perspective, lotteries seem worse than elections in ensuring accountability, in that democratic citizens facing a lottocratic legislature will have even less reason to concern themselves with political matters than they did before (Landa and Pevnick 2021). Lotteries mean that most of us will have no say in who governs us, no chance to compete for political office, and a vanishingly small chance of making politically consequential decisions ourselves. We will be poorly placed to tell whether those holding randomly selected legislatures to account have done so adequately, and it will be all but impossible for us to tell whether they have done so in ways that fairly distinguish the individual and collective responsibilities of assembly members, and therefore protect their equality as citizens. It is therefore difficult to see why lotteries would better secure democratic forms of accountability than elections, or how they correct for suboptimal levels of knowledge, judgement and political interest.

Nor is it clear how good randomly selected legislatures will be epistemically. Guerrero is keen to stress the ability of randomly selected assembly members to judge and make use of expert opinion impartially, compared to the partisan bias afflicting elected legislators. However, the epistemic quality of assembly deliberations and outcomes is likely to depend on the way lottocratic legislators interact with experts. Footnote 7 Should they defer to expert opinion? How should they handle expert disagreement? Should they set expert opinion to one side if they think it is mistaken? How one decides such matters is likely to affect the quality of lottocratic decisions, as of elected ones, and the ability of society to draw on expert opinion as needed.

Unfortunately, Guerrero says nothing on the subject, but these questions are as pressing for the epistemic authority of lottocratic assemblies as for elected ones. Indeed, they may be more pressing, because disagreement amongst legislatures is likely to be a greater threat to the epistemic authority of sortition assemblies than to their elected counterparts. The epistemic claims of elected legislatures do not depend on the unanimity of their decisions but on the quality of the deliberation that preceded them, and on the institutional framework in which it took place (Cohen 2009, 1993; Cohen and Joshua 1986). That is why the ability of a legislature to articulate disagreement effectively, and to render decisions that reflect that disagreement, can promote, rather than threaten, confidence in its epistemic, moral and political virtues.

By contrast, randomly selected legislators have no special claims to wisdom qua individuals—anymore than to moral or political authority. Such claims as they have are purely collective. As Peter Stone explains: ‘Elected representatives can say, we represent you because you chose us. Randomly-selected representatives can say, we represent you because we are indistinguishable from you. (The latter, but not the former, is inextricably a collective claim.)’ (Stone 2021b, p. 9). To the extent that the decisions of a randomly selected assembly depart from unanimity, then, they lose their collective character, thereby weakening the claims to wisdom, morality or prudence that can be made on their behalf. We may be convinced that our randomly selected representatives did their best—that they were suitably impartial, hardworking, sincere and attentive to the advice they received. But if their decisions are not unanimous, they simply become the opinions of one set of randomly selected volunteers rather than another. As Lafont has emphasised, in those circumstances random selection gives us no reason to favour the majority over the minority morally, politically or epistemically (Lafont 2020).Footnote 8

Despite Guerrero’s confidence, then, it is difficult to see why lotteries would lead to epistemically better decisions than elections—at least in the case of legislatures. These difficulties are compounded given the reasons to doubt that those people willing and able to serve, if selected by lot, are any more likely to reflect the experiences, and therefore the knowledge, of most citizens than elected legislators themselves. Guerrero hopes to make the prospect of serving in a randomly selected legislature as appealing as possible, and therefore to make it more likely that the formal equality secured by unweighted lotteries results in a legislature that is as diverse as the population from which it is drawn. However, he pays no attention to the epistemic and moral burdens that go with deciding on behalf of millions of others, when up to that point, one may have found the burdens of being a parent, a small employer, an employee, a teacher, even an administrator or local politician quite demanding enough. An impressively large salary, and help locating and relocating one’s family (p. 170) and other important material supports may simply not be enough to make such burdens bearable, let alone attractive to citizens if, as is likely, those selected can readily think of others—family members, co-workers, friends and acquaintances—who would better meet the moral, political and epistemic challenges of being a representative than they. It is therefore hard to believe that citizens’ willingness to take up their legislative opportunities will be a dramatic improvement over the very low rates of voluntary participation that currently characterise most randomly selected advisory bodies (Jacquet 2017, 2020). These range between just under 4% for the larger citizen assemblies and to average 15% across them all and explain Philippe Van Parijs’s surprise that it could take 50,000 phone calls to create a body of about 700 people for a G1000 in Brussels. Footnote 9

Very low rates in the willingness to serve threaten the epistemic qualities of randomly selected bodies, implying that members are likely to be more similar to each other than they are to the population as a whole. Unfortunately, that problem does not go away if we replace unweighted lotteries by weighted ones, thereby creating an assembly that looks more descriptively representative than would one based on unweighted lotteries alone. Using weighted lotteries means that we cannot give everyone the same chance of being selected for legislative office, but it does mean that we can create a more socially diverse legislature than if we give the same chance to be selected to all social groups regardleess of the likelihood that their members will be able to serve, even if they wanted to.

There are compelling moral, political and epistemic reasons to prefer weighted to unweighted lotteries, and their use is common practice in citizen assemblies, though the bases for weighting are not always the same (Lever 2023). However, assemblies created by weighted lotteries are not therefore more diverse epistemically than elected assemblies, nor sufficiently diverse to justify a general preference in their favour on epistemic grounds. The problem in part is the relationship between epistemic diversity and the characteristics—such as age, sex, education, profession, income, geography—used to sample the population; and, in part, the difficulty of attributing representativeness to a tiny group of volunteers from a very much larger population, even if they were also selected by lot.

The first problem arises because we cannot read women’s beliefs about the political or personal significance of their sex and gender off the fact that they are women rather than men. As Melissa Williams maintained, ‘no defensible claim for group representation can rest on assertions of the essential identity of women or minorities; such assertions do violence to the empirical facts of diversity as well as to the agency of individuals to define the meaning of their social and biological traits’ (Williams 1998, p. 6) (see also Spelman 1988). In particular, if we want to represent the epistemic differences amongst women, we need to ensure that the way we sample women, as distinct from men, adequately reflects factors, such as religious or gender differences, that are likely to illuminate the epistemic differences and disagreements amongst women. Failure to do so means that our weighted lotteries will fail adequately to reflect the ways that women can differ from each other while still being women, rather than men (Lever 2022, 2023).

The second problem arises because weighted lotteries do not change the fact that sortition assemblies rely on volunteers, and the willingness to volunteer for service is no higher in the case of randomly selected bodies than elected ones. In consequence, our legislative representatives are likely to be more similar to each other than they are to the general population, however they are selected—perhaps because they value legislative service more highly than other people, or because they are more self-confident. The difficulty for Guerrero’s argument, then, is that even when a randomly selected assembly reaches a decision unanimously, that decision may be no better epistemically, and no more reflective of our beliefs as a population, than if it had been taken by an elected body. Our legislatures can fail us epistemically, as well as morally and politically, by insisting on their partisan disagreements at the cost of interests we have in common. But they can also fail us by overlooking the reasons why we disagree, and the depths and intractability of that disagreement. The fact that randomly selected legislatures are likely to be made up of people who are more similar to each other than they are to the rest of the population makes such errors more likely. It is therefore impossible to predict that they will be better epistemically than elected assemblies.

Of course, we could get rid of voluntary participation and treat legislative service as though it were a form of jury service or military service.Footnote 10 But while that might improve the epistemic quality of sortition assemblies, the prospects of justifying such a move on democratic principles look unpromising. Turning the right to participate in collective decisions into a duty to serve turns democracy into a painful chore that must be justified fairly. It is hard to see how that answers to the idea that democratic government is valuable and an expression of our interests in agency as well as equality. Epistemically, it does not look promising either, in so far as it encourages people to lie about their conscientious convictions, competence or the burdens that legislative service would pose for them. Politically, compulsory legislative service is unlikely to work—or, at any rate, to solve the political problems that trouble Guerrero. Thus, moving from voluntary to mandatory legislative service is unlikely to make sortition more attractively democratic than elections.

Must Elections be Epistemic Disasters?

We have seen, then, that randomly selected legislatures are unlikely to have the properties that justify Guerrero’s faith in their epistemic superiority to elections. Indeed, his confidence in their superiority seems to rest as much on the dismal picture of electoral democracy that he draws as on the supposed advantages of sortition to election. That dismal picture of democracy is based almost solely on American evidence, and on epistemic critiques of democracy, such as Jason Brennan’s (Brennan 2016), (Achen and Bartels 2016) that seem remarkably one-sided, indifferent to the causes of American pathologies and to the prospects of remedying them (Bagg 2018). It is doubtful, then, that rejecting Guerrero’s optimism about lotteries commits us to accepting the epistemically dire alternative he presents.

Following Brennan, Guerrero claims that ‘hyper-partisanship’ (p. 166) means that we end up treating our political parties as sports teams, and rooting for them in ways that are irrational individually and collectively (p. 166). Guerrero complains that ‘[G]roup attachments and social identities drive our thinking about politics, rather than the other way around’ (p. 166). Whether this is irrational or leads to worse outcomes than ignoring them, however, depends on how far our lives are shaped by ascriptive and voluntary ties to others—and how far attention to those ties enables us to avoid misleading, manipulative or mistaken views of politics. By treating electoral rules as irrelevant to the character and dynamics of electoral politics, or as wholly determined by it, Guerrero makes it impossible to understand the significance of different electoral systems for the political opportunities, status and resources of citizens; or to see how the same electoral system can change over time in ways that promote, or impede, the epistemic diversity and the political representation of socially disadvantaged groups (compare (Mráz and Lever 2021a, b; Mráz 2021).

Even the view of partisan sporting attachments, appealed to by Guerrero and Brennan, is one-sided and condescending. People who passionately support their own team may applaud the accomplishments of their competitors; and they may deplore unfair rules or poor arbitration for diminishing the quality of competition—its ability to display participants’ skills, to engage spectators and onlookers—and not simply its consequences for who wins. Competition, then, need not be pathological, or lead to indifference to fairness, solidarity or to accountability for rule-breaking and for error. Indeed, considerations of fairness underpin arguments for proportional forms of electoral representation (Powell 2000) and, even in the USA, have supported special protections for ‘small and insular minorities’ (United States v. Carolene Products Company, 304, U.S, 144 (1938); Williams 1998). Hence, scepticism about the epistemic merits of lotteries does not condemn us to a conception of democracy as cartelised struggle on the one hand, or unhinged emotion, on the other.


The epistemic case for replacing elected lotteries with randomly selected ones is weak. But that does not mean that randomly selected citizen assemblies cannot have a variety of appealing properties if one cares about democracy. If one thinks of citizens as bearers of politically relevant knowledge,Footnote 11 randomly selected assemblies can be a helpful way to elicit that knowledge, especially because lotteries can be weighted in different ways both to test the consistency and/or robustness of citizen opinion on an issue, or to bring the opinions of particular groups into sharper focus on matters that affect them. Citizens’ lived experience is likely to make them differentially sensitive to, and knowledgeable about, the strengths and weaknesses of their collective institutions and the scope for remedying those weaknesses (Young 2002, chs 2 and 4). Their lived experience, professional habits and training may enable them to perceive the morally salient features of situations quickly and effectively in ways that would be less obvious to others (see Jakob Elster 2023). Given that randomly—selected assemblies can take many different forms, their flexibility makes them an appealing tool for securing public interest in, knowlege about and responses to public policy.

As such, randomly selected citizen assemblies constitute a welcome addition to the repertory of democratic politics. Citizen assemblies can be dispersed geographically in ways that facilitate political participation by citizens throughout the country—this, for example, was an attractive feature of the climate assemblies in the UK, as compared to the single, Paris-based Citizens Convention on the Climate, in France. Although it is customary for citizen assemblies to report their findings as a group, rather than issuing majority and minority reports, clarity about the depth and reasons for citizen disagreement can be as important to good policy-making as an understanding of the scope for agreement (Lafont 2020; Ingham 2013). Hence, experimenting with different forms of reporting may be as beneficial as experimentation with the ways that random selection proceeds, or that citizen deliberation is organised.

Citizen assemblies are used for a variety of different purposes. Some assemblies are extremely local and practical, others are explicitly national (OECD 2020, pp. 121–148). Their rationales are rather different too. Sometimes, they seek to tap local knowledge and expertise, and to facilitate responsiveness by local politicians to the needs of their tax payers and electors, as with the greater Cambridge, UK, citizens’ assembly.Footnote 12 At others, appeals to citizen deliberation are an effort to break a political deadlock where legal reform is desirable, but where politicians are unwilling or unable to anticipate public opinion for fear of being branded morally insensitive or worse. Ireland’s assemblies on abortion and gay marriage are examples of this, and, as Gray suggests, citizen assemblies might have a similar beneficial role on gun control and universal healthcare, in the USA.Footnote 13 Assemblies can be used to anticipate future policy debates—to provide a form both of ‘alerting’ and ‘sounding out’ the public on challenges that lie ahead—for example on biotechnologies, but also in the rationing or distribution of scarce, but important resources such as healthcare (Littlejohns and Rawlins 2009). Freed of electoral incentives, they can provide the long-term perspective that it is hard for politicians to adopt (see Philippe van Parijs in Van Reybrouck et al. 2014) and, if numerous enough and persuasive enough, can create the electoral pressure that politicians need (or seek) in order to take difficult decisions.

The democratic value of these deliberative experiments does not depend on agreeing to a set of ‘good practice principles’ (Bellantoni et al. 2020),Footnote 14 rather than the ability to relate the principles used in their construction to the deliberative purposes for which they were sought. At present, citizen assemblies pay insufficient attention to the challenges of integrating disadvantaged groups into deliberation, and recognising that their perspective may be grounds for rethinking how deliberation is organised, and what questions and expertise are used to structure it. While well meaning, for example, the effort to secure two places for those in ‘deep poverty’ in the French climate assembly was sheer tokenism absent efforts to ensure that they could effectively articulate what climate change, and different strategies to combat it, might mean to them (Fabre et al. 2021).Footnote 15 If it matters that an assembly understands what climate politics look like to the most disadvantaged, their views must have the possibility of framing discussions, not simply contributing to debates and decisions already established by others. And as a careful qualitative study of citizen deliberation reveals, role-playing and other departures from idealised conceptions of deliberation may help assembly members to assimilate the significance of the information they are receiving from experts, and from each other (Davies et al. pp. 129, 194–195, 92). Thus, rather than trying to standardise the ways in which assemblies are constructed and operate, it may be better to be open about the advantages and disadvantages of contrasting approaches to assembly construction and organisation, including trade-offs between more intensive small-group discussion and the goal of inclusive deliberation by a broadly representative part of the population (Lever 2013, pp. 103–105).

In short, there is no need to turn citizens into experts—with special access to moral knowledge or special claims to epistemic and political authority—to suppose that they are entitled to take part in public affairs not simply as political partisans, but as people with no fixed political identities and loyalties. If Guerrero is wrong to suppose that the worst features of the American political system are an inevitable consequence of using elections rather than lotteries to create legislatures, contemporary democracies could benefit from more appealing, inclusive and deliberative forms of citizen engagement than are currently available. As Philippe Van Parijs insists, the limitations of citizen assemblies as instruments of democracy, ‘make modesty inescapable’(van Parijs 2014, p. 52) in our claims for them. But that is fully compatible with them being important additions to the deliberative and participative repertoire of electoral democracies, along with referendums and other, more directly democratic, forms of politics.