What Deliberative Democracy Actually - IsThe Available Remedies, Part Two

The Available Remedies has committed to testing each proposal against the upstream principle before either recommending it or declining to. The first remedy to be placed on that test is the one most often offered as the answer to the conditions the diagnostic series named: deliberative democracy, citizen assemblies, sortition. The concept arrives with a literature, a growing base of advocates, and a track record substantial enough that treating it as a democratic inheritance contemporary government has allowed to lapse is no longer an eccentric claim. It deserves examination on its merits rather than adoption on its enthusiasm. The same upstream test is applied here as has been applied elsewhere in the series to existing arrangements. A reform that cannot withstand the discipline the diagnostic work required of defenders of the status quo is not yet a remedy. It is an aspiration.

A distinction first. The term deliberative democracy covers forms of practice that do not operate at the same level of the causal chain. A citizen jury advising a local council on waste management, and a constitutional assembly drafting amendments for national referendum, are both instances of deliberative democracy in the academic literature, but they are not equivalent in their relationship to the structural conditions this series has diagnosed. One is a refinement of existing consultation. The other is a restructuring of how binding decisions are taken. The upstream principle does not ask whether deliberative democracy is valuable in the abstract. It asks, at each application, whether the practice in question intervenes at the level where the condition lives.

Walker Briefing is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

That distinction governs everything that follows.

The Remedy, Briefly

A deliberative assembly is a body of citizens selected by stratified random sampling to approximate the demographic composition of the population, brought together for an extended period to consider a specific question, given expert briefing from multiple perspectives, and asked to produce a recommendation or, more rarely, a decision. The method has a pedigree of roughly two and a half thousand years. Classical Athens administered most of its civic offices by lot, including its juries, its executive boards, and its Council of Five Hundred. The method was not peripheral to the Greek democratic experiment. It largely defined it.

The modern revival has been gradual. Danish consensus conferences developed through the 1980s, the British Columbia Citizens’ Assembly on Electoral Reform ran through 2004, the Irish Constitutional Convention followed in 2012 and then the Citizens’ Assembly of 2016 to 2018, the French Convention Citoyenne pour le Climat through 2019 and 2020, the Mongolian constitutional deliberation in 2017, and across that same period the Stanford deliberative polling programme under James Fishkin in more than one hundred deliberations across twenty-eight countries. The method has been extensively studied. The findings are consistent enough that they can be stated without qualification. When ordinary citizens are given an important question, access to evidence, structured facilitation, and adequate time, they perform well. Participant enthusiasm is near-universal. Outcomes tend to shift toward the centre, partisan divides soften, and participants emerge better informed than the electorates from which they were drawn.

John Macgregor, in the book this series engages as its primary resource on remedy, summarises the case as follows. Deliberative democracy provides, in his formulation, “a clear indication of what the considered judgement of the entire adult population would be if they were able to deliberate on issues thoroughly, freely, and in an informed manner.” That is a strong claim. It is also a claim that moves quickly from what deliberative bodies demonstrably produce within their own walls to what they can be relied upon to deliver at the level of binding public decision. The gap between those two things is where the rest of this piece lives.

The Evidence, Honestly Read

The strong evidence first. The Irish Citizens’ Assembly of 2016 to 2018 brought ninety-nine citizens, two-thirds of them randomly selected, to consider a range of constitutional questions including abortion, fixed-term parliaments, climate change, and population ageing. It cost approximately 2.4 million euros. It produced eighteen recommendations for constitutional change and twenty for legislative and procedural reform. The 2018 abortion referendum, taken to the electorate on the Assembly’s recommendation, passed by 66.4 per cent. The 2015 marriage equality referendum, which emerged from the earlier Constitutional Convention, passed by 62.1 per cent. In both cases, proposals generated by deliberation were endorsed by a wider electorate on terms that the deliberative participants themselves could not have guaranteed.

Professor James Fishkin and Professor Larry Diamond assembled the largest deliberative body in American history in 2019. Five hundred and twenty-three citizen delegates deliberated on five highly contested issues including immigration, taxation, health care, and the environment, with a bipartisan expert panel addressing their questions and five presidential candidates participating. Most participants moved toward the centre across all five issues. The proportion believing the other side had good reasons for its views rose from thirty-four per cent to fifty-four per cent during deliberation, with most of the attitude shifts still present a year later.

Mongolia convened a deliberative body of six hundred and sixty-nine randomly selected citizens in 2017 to draft amendments to the national constitution, with technical support from the Stanford Center for Deliberative Democracy. The resulting amendments were adopted by the Mongolian parliament in 2019. This is the clearest contemporary case of deliberative democracy operating at the level of constitutional architecture and producing implementation.

Ostbelgien, the German-speaking community of Belgium, has gone further than any other jurisdiction. In February 2019, its parliament voted unanimously to establish a Permanent Citizens’ Dialogue, the first deliberative mini-public institutionalised through legislation and given a continuing role in the legislative process. A twenty-four member Citizens’ Council sets the agenda, selects topics, convenes short-term Citizens’ Assemblies of up to fifty people drawn by lot, and monitors the follow-up. Parliament is legally required to debate the recommendations and to respond publicly within a defined period. By early 2025, six Citizens’ Assemblies had been convened, most of which had gone through the follow-up process the founding decree requires. The population of Ostbelgien is approximately eighty thousand. The scale matters. Nonetheless, the model establishes the proposition that citizen deliberation can be given standing rather than contingent status, and that parliament can be required to respond rather than being free to choose whether to engage.

A reader who dismisses this body of evidence is not engaging with the record. Deliberation, at the level of deliberation itself, does what its advocates claim. It produces considered judgement from ordinary citizens confronted with evidence, softens partisan reflex, and generates outcomes the wider public can recognise as reasonable. That finding is not in serious contest in the academic literature.

The Harder Evidence

The same cases, examined at the point where deliberation meets implementation, tell a more qualified story.

The British Columbia Citizens’ Assembly on Electoral Reform deliberated through 2004. One hundred and sixty members, one man and one woman drawn at random from each of the province’s seventy-nine electoral districts, plus two First Nations members and a chair. The Assembly heard expert evidence, consulted the public, and in October 2004 recommended adoption of a single transferable vote system to replace first-past-the-post. The recommendation went to referendum the following year. It received 57.7 per cent of the vote, including majorities in seventy-seven of the seventy-nine electoral districts. Because the provincial government had unilaterally established a 60 per cent super-majority threshold, the reform did not proceed. A second referendum in 2009 produced 39 per cent support. The Assembly’s work, by every measure of deliberative integrity, was sound. It was defeated at the translation point, not at the deliberative one.

The French Convention Citoyenne pour le Climat ran from October 2019 to June 2020. One hundred and fifty randomly selected citizens produced one hundred and forty-nine proposals after nine months of intensive deliberation. President Macron had committed to submitting the proposals “without filter” to referendum, parliament, or direct executive implementation. Within a week of receiving them, he used what he called three “jokers” to reject three of them outright. The Climate and Resilience Law that emerged in July 2021 was, on the assessment published by the Deliberative Democracy Digest, a substantially diluted version of the Convention’s work. Ten per cent of recommendations were accepted without modification, thirty-seven per cent were modified or watered down, and fifty-three per cent were rejected. The promised constitutional referendum was abandoned. The deliberation was, by every participant’s account, genuine. The translation into policy was not.

The Irish cases, most often cited as the strongest example of deliberative democracy succeeding, are themselves more mixed than advocates typically concede. Government decided which recommendations reached referendum and which were quietly set aside. The assemblies produced the deliberation. They did not control the selection of what became binding. There is a further point rarely made in the Irish literature. The referendums carried in Ireland on marriage equality in 2015 and abortion in 2018 concerned questions on which the Irish public appears to have done a great deal of underlying work before the assemblies convened. The assemblies performed well, and on the available evidence they formalised a judgement the public was already substantially prepared to make. What the Irish record does not yet demonstrate is that an assembly can create readiness where the ground is cold.

The pattern across these cases is not that deliberative democracy fails. It is that the deliberation works and the translation does not automatically follow. Every policy process, of course, depends on implementation architecture to some degree. The specific finding here is sharper. Deliberative democracy is being advocated as an upstream remedy for the structural conditions this series has diagnosed, when in the configurations currently in circulation it operates downstream of the same architectures that produce those conditions. The decisive variable is not the quality of the deliberation. It is the architecture of the translation mechanism that governs what happens after the deliberation concludes. This is the central finding of the examination, and it is worth naming as such rather than leaving the reader to infer it.

The South Australian Counter-Example

One Australian case cuts against the pattern, and it is worth attending to because it specifies the conditions under which the remedy can bite.

In 2015 the South Australian government established a Royal Commission into the Nuclear Fuel Cycle. The Commission reported in May 2016 and recommended that the state pursue the establishment of storage and disposal facilities for multinational high-level and intermediate-level nuclear waste. The projected revenue to the state was estimated at a substantial figure. The government convened two citizens’ juries to deliberate on the Commission’s recommendations. The first jury, meeting in June and July 2016, structured the terms of state-wide consultation. The second, comprising 350 randomly selected South Australians and meeting through October and November 2016, was asked to consider under what circumstances, if any, South Australia should pursue the proposal.

The architecture was, by design, a consultation intended to generate assent. The Premier had endorsed the Commission’s findings. A pro-nuclear information campaign had toured the state. Two-thirds of the second jury rejected the proposal under any circumstances. The jury’s report questioned the Commission’s economic modelling, noted the absence of Aboriginal consent as a foundational obstacle, and refused the framing the consultation architecture had established. The Premier subsequently proposed a state-wide referendum, which did not proceed, and in June 2017 announced that the plan would not be progressed.

The case is instructive on two levels. The citizens declined to do what the consultation was built to have them do, demonstrating that a deliberative process honestly facilitated is not guaranteed to produce the output the commissioning body anticipates. The outcome was implemented because the political class, confronted with the jury’s refusal and the wider public reception of it, found the project no longer viable. This is the difference that matters for the analysis. Ostbelgien illustrates designed translation, in which the architecture is built to carry deliberative output through to parliamentary response. South Australia illustrates contingent alignment, in which deliberative output reached an implementation point only because it coincided with an already-mobilised political environment. Both are successes. Only the first is reproducible on demand. The distinction matters because proposals that rely on alignment rather than architecture cannot be designed for. They can only be hoped for.

The Structure of Assembly Control

Before the remedy can be said to have been examined against the upstream principle, the same structural scepticism this series has applied to existing institutions must be applied to the assembly itself.

In every case currently in the literature, someone chooses the question the assembly is asked to deliberate. In Ireland that someone is the Oireachtas, in France the executive, in British Columbia the provincial legislature. The framing of the question is a structural act that precedes and constrains the deliberation. Someone selects the experts who brief the assembly. Typically a secretariat, typically reporting to a minister or ministerial committee, making decisions that shape the evidentiary terrain on which the deliberation proceeds. Someone drafts the options the assembly works through, generally the same secretariat, operating within parameters it has not itself been asked to justify. Someone controls the timeline, which, because deliberation is expensive, effectively means the commissioning body. And someone holds the recommendation after the assembly finishes, which in every case except those with a constitutional referendum requirement means the existing political system, and even there, as the Irish experience shows, government decides which recommendations reach the ballot.

Each of these control points can be answered in ways that make the assembly a genuine upstream restructuring of deliberation, or in ways that make it a sophisticated downstream instrument providing cover to existing decision-makers. The architecture determines which. The name “citizen assembly” does not. An assembly whose question, experts, options, timeline, and translation mechanism are all governed by the system it is supposedly reforming will not reform that system. It will be absorbed by it.

A reader may notice that this argument, pressed hard, threatens to collapse the entire remedy class. If existing institutions control every entry point to the assembly, then the assembly is necessarily captured, and no reform of this kind can ever be genuinely upstream. The Ostbelgien case shows why this collapse is not automatic. The design located topic selection, follow-up monitoring, and parliamentary response obligations with the Citizens’ Council itself rather than with the executive. Architecture can be designed to shift the control points, and the architectural choices that matter are now sufficiently well-documented that they are available to Australian designers if the political conditions to adopt them develop. What cannot be done is to import the assembly mechanism without importing the control-point redesign, and to expect the result to behave as the Irish or Ostbelgien models do. On current Australian evidence, that is precisely what most proposals in circulation appear to assume.

The Australian Conditions Test

The most instructive recent Australian evidence on this class of proposal is not strictly a citizen assembly, but it illuminates the translation problem the series has been tracking.

The Uluru Statement from the Heart emerged from the regional dialogues of 2016 and 2017 and the convention at Uluru in May 2017. That process was a deliberative architecture, conducted among the Indigenous Australians directly affected by the proposal it produced, operating according to recognisable deliberative principles and generating a coherent institutional proposal with legitimacy within the deliberating group. The 2023 referendum asked the wider Australian electorate to endorse the resulting institutional proposal. The national Yes vote was 39.94 per cent. No state returned a Yes majority.

The Australian National University’s post-referendum study by Biddle, Gray, McAllister and Qvortrup recorded a finding that matters here. There was no majority opposition to constitutional recognition of Indigenous Australians in principle, and strong majority support for Indigenous voice in matters affecting Indigenous Australians. It was the specific institutional proposal that was rejected. The vote did not signal a lack of support for reconciliation, truth-telling, or pride in First Nations cultures, each of which polled at around eighty per cent in the same study.

What the Voice process illustrates, for a series examining deliberative democracy as a remedy, is not a simple failure of that class of process. The Uluru architecture was deliberative within the group it was designed to include, and the output that process generated had legitimacy within those terms. What the 2023 referendum tested was something the architecture had not been built to deliver: the transmission of deliberatively-derived legitimacy across a membrane that the deliberation had not included, to a wider public without its own deliberative engagement with the proposal. A compressed national campaign was expected to do translation work that comparable Irish proposals achieved through an extended sequence of Oireachtas committee engagement and public engagement. The Irish arrangement built translation architecture. The Australian arrangement substituted campaigning for it. The argument here is not that better translation architecture would have secured the referendum’s passage. That is a counterfactual this analysis cannot settle. The argument is that a known structural gap was left unfilled, and that the outcome was consistent with what the upstream principle would predict when that gap is left unfilled. Campaign dynamics, trust, and political messaging all shaped the specific result. What the framework isolates is the layer relevant to its own thesis.

What Western Australian Conditions Specifically Suggest

Three conditions in this state would have to be addressed honestly before a proposal for deliberative institutional reform could be considered upstream rather than another centralised consultation conducted from the capital. What follows is design inference based on observable features of Western Australian civic life, rather than tested evidence from implemented trials. It should be read as an identification of the questions any proposal would face, not as an answer to them.

Participation is the first question. Deliberative architecture depends on sustained engagement, and a full week away from home is not feasible for a fly-in-fly-out mining worker on roster, a pastoralist during mustering season, or a remote community representative three days’ travel from Perth. Hybrid models of local deliberation linked to state-level integration are technically possible and have been piloted at municipal scale elsewhere, but the design work required to make them function across Western Australia’s geographic dispersion, with its attendant cultural and linguistic diversity, has not been done.

Institutional integration is the second. The state already operates a range of mechanisms for gathering non-expert judgement on public questions, including royal commissions, parliamentary committees, legislative reviews, community cabinet meetings, and formal stakeholder consultations. A citizen assembly is either a structural addition with defined authority, a replacement for one of these mechanisms, or a parallel process risking polite reception and quiet disregard. The choice must be made and defended at the outset. An assembly established without a clear relationship to the institutional architecture it joins is an assembly whose recommendations that architecture can absorb or reject at its convenience.

Translation is the third, and it is the question the Voice has placed before any Australian proposal. Between an assembly’s output and any binding decision, what architecture performs the translation work? A parliamentary committee with standing to hold the recommendation? A statutory requirement for substantive government response within a fixed period? A citizen-initiated confirmation referendum? Each is available in principle, and each is absent from the existing architecture. Without a mechanism deliberately designed to carry deliberative legitimacy across the membrane between the assembly and the wider public, an assembly can deliberate faithfully and produce a coherent recommendation, and the system can decline to adopt it on terms the system controls. That is the French outcome. That is the British Columbian outcome. Australia has no structural reason to suppose its outcome would be different in the absence of architecture designed to be so.

Where the Examination Stands

The remedy is genuinely promising under certain architectures and genuinely hollow under others. The same name covers both cases. An article that recommends deliberative democracy without specifying which architecture is advocacy. An article that dismisses it without engaging the successful cases is not examination. What the discipline of the series requires is the more uncomfortable intermediate position: the evidence supports the remedy at the level of deliberation, problematises it at the level of implementation, and makes the architecture of translation matter more than the architecture of the assembly itself.

There is a further question this piece has only pointed toward, which the next article in the series will take up directly. An assembly deliberating honestly cannot answer the question of what it has been given to deliberate about. The choice of question is made upstream, and the choice is itself a structural act. If the question handed to the assembly is already calibrated to the incentives of the existing political system, faithful deliberation will still produce an outcome shaped by those prior incentives. That is the territory of the next piece, on what the financial architecture of the political system does to every remedy, including this one, before the remedy has a chance to do its work.

The patient has been told what the first remedy is. The question of whether to take it is specified. It is not yet settled.

This article is Part Two of The Available Remedies, a series on the Walker Briefing examining structural responses to the conditions diagnosed in the preceding Zhōng Yōng series. The full archive is available at bfwalker.substack.com.

Walker Briefing is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.