Collective Reason, the Rationality Gap, and Political Leadership

The article discusses the implications of the well&#8208;known discursive dilemma. The dilemma arises whenever a reasoned decision has to be taken by a collective decision&#8208;maker and generates persistent contradiction between what is defined as collective reason and public opinion. Following Philip Pettit, I argue that collective reason is normatively preferable and that the role of existing constitutional institutions in contemporary democracies is to collectivise reason. However, this makes the frustration of popular will a systematic by&#8208;product of any well&#8208;functioning political process. I argue that the only way out is if individual beliefs are subject to revision during cycles of public deliberations with the responsibility to lead this thrusted upon elected representatives.

apparent than real. This is even more striking given the abundance of literature on what the representatives, once elected, should do. There is also considerable interest in various nonelective forms of representation, but all of these are complementary to electoral representation, the existence and necessity of which is hardly ever questioned. 1 In the age of Cambridge Analytica, this consensus begins to crumble. Studies show that younger generations do not consider democracy and elections as important as they are for their parents and grandparents. 2 Books announcing the end of democracy become academic bestsellers. This article aims to contribute to this debate by showing that the current malaise is not a contingent failure of certain institutions, policies, or people but that electoral democracy breeds its own opposition on a systemic basis. My point of departure is the claim that the proper inputs in the political process are not people's policy preferences but their beliefs about certain policy-relevant facts, i.e., about certain reasons for action. The right collective action is to be determined on the basis of these beliefs, yet the relationship is a complex one and these actions are not a function of these beliefs. If this argument is correct, the majority will is systematically frustrated by what is to be defined as collective reason. This is by no means a new phenomenon, but it has become visible only recently. I argue further that elected representatives are both the problem, as they have every incentive to exacerbate the tension, and the solution, as they are the only ones who are able to provide one. I posit that their proper function is that of mediation between certain inputs in the political process and certain outputs thereof.
The article starts with a discussion of a well-known problem of democratic decision-making to show that while collective decisions must take democratic inputs into account, the ultimate decisions should not be responsive to them (Section 2). Then I shall address the tension this creates (Section 3) and shall suggest that elected representatives can and ought to resolve these tensions, to conclude that elections are instrumental in making them responsible to do so (Section 4). The latter is, in my view, difficult but possible, as elected representatives possess both the ability to track the beliefs held by citizens and the ability to shape these beliefs (Saward 2006). In the final two sections, I shall discuss two ways democracy degenerates when governments fail to carry out their responsibility to resolve the tension. These are technocracy (Section 5) and populism (Section 6) respectively.

Responsiveness and Responsibility
Lincoln's definition of democracy as "government of the people, by the people, for the people" provides as good point of departure as any, and Fritz Scharpf (1998) famously elaborated on the meaning of the latter two components. According to 1 See, for example, the entry in the Stanford Encyclopedia of Philosophy, which provides comprehensive and up-to-date review of the literature (Dovi 2018). Dovi notes that due to the emphasis on elections, discussions about the concept of political representation frequently collapse into discussions of democracy and references Schumpeter, who considers competitive elections as the criterion by which to distinguish democracies from the rest. 2 One survey made headlines around the world by finding that 46% of Americans aged 18 to 29 would prefer to be governed by experts (Wike et al. 2017). Another study found that a quarter of the same group believes that "choosing leaders through free elections is unimportant" (Foa and Mounk 2017) and the authors claim that results are similar across the globe. has to decide on a certain policy to deal with irregular immigration, for example by building a wall along the border with a certain southern neighbour. Suppose also that among them it is undisputed that there are three relevant premises the cumulative availability of which warrants the approval of the policy. These are P 1 -immigration threatens jobs and public safety; P 2 -building a wall will stop or at least decrease immigration; and P 3 -the cost and timeliness of the construction are feasible. However, the members' beliefs on each of the premises differ as follows: Members P 1 : Immigration a problem? Apparently, there is a majority believing that each of the premises obtains. Therefore, if the panel decides by voting on each of the premises, the decisions on all three will be positive, and therefore the construction of the wall should be deemed approved. However, if the panel does not decide on the premises separately, but each of the members makes up her own mind whether to give the go-ahead, there is a majority against the construction. The result, in the bottom-right cell of the matrix above, differs depending on whether the decision is taken in premisewise mode (PWM) or in conclusionwise mode (CWM). Certainly, the actual occurrence of this contradiction depends on the distribution of the members' judgements on the premises (and on the set of premises which are relevant); however, in the complex policy-making of today distributions where the contradiction does arise would be common. It is important to note that CWM decision-making is apparently responsive to the opinions of the decision-makers (a simple aggregation thereof); in technical terms, it is a function of their beliefs on the consequences (which translate into their policy preferences). However, even though in PWM the decision is also fully determined by the beliefs of the decision-makers, it is not a function of these beliefs. Finally, it should be obvious that "collective rationality," in the sense it is used here, is quite a different concept from "rationality"-what is the collectively rational thing to do depends on the actual beliefs of the panel members, and of the people, and may be quite irrational from an external perspective.
The original paradox was identified in cases of jury trial where the premise-conclusion relationship was specified by mandatory legal rules (wrongdoing, intent, etc.), but the result can be generalised: "a paradoxical distribution of views can arise in any group of three or more persons faced with a decision that can be broken down into at least two constituent sub-decisions" (Kornhauser and Sager 2004, 251). Even if the decision-makers are not formally bound to follow any doctrinal rule, the reasons they give to publicly justify their decisions are normally expected to exhibit a measure of consistency, i.e., to conform to the principle of integrity. 3 Thus formal logic, common sense, or recognized scientific laws may determine the set of relevant premises and bind the decision-makers just as legal doctrine binds courts. This is more interesting than it may appear, as rule-makers are typically considered free in their choices (within the constitutional limits). Nonetheless, they often face the same integrity challenge. In other words, they will feel under pressure to collectivise reason.
Thus, whenever social choices are explicitly made on the basis of reasons, 4 it is likely that various majorities support each of the reasons, yet a majority opposes the conclusion that logically follows. 5 Again, the probability of this contradiction actually occurring depends on the distribution of the members' beliefs. This is an empirical question and beyond the scope of this paper. It suffices to say that in contemporary polities hundreds of decisions are taken on a daily basis and that it is quite rare for a policy decision to be taken in isolation. Any policy must be considered with regard to its effectiveness and financial underpinning (P 2 and P 3 ) as well as its health and safety, its implications for the environment and for national security, respect for human rights, etc. While the discursive dilemma may occur whenever decisions are based on two or more logically connected propositions, in practice policy decisions are based on dozens of premises. Although no empirical polity follows the 3 "A system satisfies the principle of integrity if it supports are consistent with each other" (List and Pettit 2005, 378). Consistency is judged according to logic or some other set of external rules. 4 Following Scanlon (1998), I define reasons as considerations that count in favour of doing (or abstaining from doing) something. For the sake of simplicity in this article I shall use reasons, considerations, premises, and beliefs on premises interchangeably. 5 Note that CWM enables many "incompletely theorised agreements," as it allows each member to support a conclusion for different reasons. PWM is instead likely to expose some unholy reasons and disrupt such agreements. On the other hand, PWM may induce people who disagree on the conclusions but share beliefs on some of the relevant premises to iron out the remaining differences and reach agreement. This might be especially the case when the judgement on certain premises is dependent on input from a trusted source.
Collective Reason and Political Leadership premisewise procedure described above, below I shall show that the relevant premise-beliefs are made explicit in a number of formal and informal ways, thus exerting pressure on the decision-maker to decide as if they were following the PWM model. As Kornhauser (2011) notes, such a dilemma may occur whenever a collegial decision-maker is required to give reasons for their decision.
Note that this will hold also if the decision-making body is extended beyond the three members to the whole population, or, more interestingly, to any kind of authority which takes into account the relevant premise-beliefs of the population. 6 It is worth noting that sometimes all members of the panel may individually oppose the policy even if it is collectively rational to have it adopted (in the example above this will be the case if A judges P 2 to be true but B judges it to be false). Whenever this is so, the panel will have the two opposite views at the same time-a collective opinion in favour of the policy and "personal" views of all members against it. 7 Thus far we have taken the set of reasons on which the decision is premised for granted. Of course, in reality the relevance and also the weight of each premise will be highly contested. For example, it may or may not be considered what is the effect of immigration on the crime rate or on the established traditions of the host society. What will be taken into account depends on the particular constellation-constitutional law may determine whether the religion of the incomers may be taken into account, experts (or elaborate impact assessments) may say whether immigration affects employment at all, 8 or the public-lay citizens, or political entrepreneurs, with the help of the media-may make salient considerations which were previously irrelevant (e.g., immigrants' country of origin). Thus, it must be noted that the decisional matrix is not determined a priori and is rarely taken for granted; nor is it set by the elites only. That is why premisewise decision-making should not be mistaken for technocracy. Even if premises are proposed by the experts-which is the case in a number of environmental or financial issues-their inclusion and weight in the matrix and, of course, the individual judgements on each of them, are overtly political choices. I shall revisit the issue of technocracy in Section 5 below, where I argue that collectivisation of reason and technocracy are antithetical, and the latter emerges when the former fails.
The uneasy choice between the two modes of decision-making Pettit calls discursive dilemma and notes that going the conclusion-driven way means adopting a course that is inconsistent with the premises endorsed by the group and going the premise-driven way means adopting a course that a majority individually reject. Going the first way means sacrificing collective rationality for the sake of responsiveness to individuals, going the second means sacrificing responsiveness to individuals for the sake of collective rationality. (Pettit 2001, 274) 6 The panel members may be seen as deciding not whether P 1 -P 3 are true or false, but whether their respective constituencies or the electorate as a whole believe P 1 -P 3 are true or false. 7 This distinctive opinion that a collective may form is the ground on which Philip Pettit and Christian List maintain that groups may be agents in their own right (Pettit 2003;List and Pettit 2011). 8 As far as I am aware evidence suggests that it does not, and of course, "the people" have had enough of evidence. But both of these points are irrelevant; the point I am making is that on some issues, at a certain moment, some expert may have the final word. Whether certain expert views, or certain prejudices, will be normalized and carry the day depends on the public discourse.
There are a number of good reasons for a polity to prefer to go along the premisewise way. One is that if the act of authority is not supported by publicly supported reasons it would be arbitrary and illegitimate, at least on a republican account of legitimacy (Pettit 1997). The other is that the collectivity will fail in terms of practical reason-it will not be able to set certain goals and systematically take decisions which are instrumental to achieving them. The polity will act inconsistently and successive decisions will cancel each other. 9 That is why Pettit unequivocally prefers PWM even though he acknowledges that in this way responsiveness to individual wills (i.e., to their aggregation into a "popular will") will be lost; for him this is a price worth to be paid for the gains in collective rationality. In choosing that way Pettit finds himself in good company-think of the Federalist Papers, for example-but his reason for it is his republican understanding of freedom as independence from arbitrary power (Pettit 1997). Apparently, a CWM decision is arbitrary, for it is not premised on any reasons which are collectively supported; therefore, it is based on no reasons at all. Now, one does not have to be a republican to strive for collective rationality. Even traditional liberals may be unable to satisfy themselves with the conclusionwise way, because in many cases the reasons for these acts are not altogether irrelevant for citizens. Thus, the government administration everywhere is held to account, i.e., it is required to explain and justify its actions in parliaments and in the media; it may be judicially reviewed, including for the adequacy of its reasoning. Although legislatures are freer, they are also increasingly required to accompany bills with explanatory memoranda, base them on impact assessments, etc. In short, justification matters.
One still may argue that justification is cheap talk, and any half-awake minister will be able to come up with reasons for any decision ex post. Yet when the authority is a collective body, reason-giving becomes tricky. Certainly, it may very well jump to voting on the conclusions without any deliberations at all, yet if it has to produce some reasons it will soon run into the discursive dilemma too. 10 In our example above, if the panel decides conclusionwise, it will find itself in a position to state reasons each of which is disbelieved by a majority of its members. Certainly, the members have the power to reject the measure as per their individual preferences, but will not be able to state that P 1 , P 2 , or P 3 does not obtain, because on each of these a majority thinks otherwise. Of course, language is fuzzy and the political process often tolerates ambiguities, so such a majority may still be imaginative enough to produce some more or less vacuous statement that can be agreed by all members. Yet the very fact that they will have to do so suggests that the panel is under some degree of pressure to collectivise reason. Those who see representatives as mere conduits for the preferences of the people may still want to deny that this represents a problem. Indeed, democratic decisions are expected to reflect the wishes of the majority regardless of the reasons citizens may have for these wishes, and of course citizens do not have to give any account for their reasons to prefer one policy to another. Still, it is hardly acceptable for a prime minister today, when asked to give an account in parliament for a certain policy choice, to answer merely "because so we chose" or "because most people think so." She would rather go out of her way to show that this choice somehow follows from previous choices and commitments and that it is appropriate and beneficial for "the people" as a whole. Decision-making on levels lower than ministerial are subject to an even denser web of mandatory considerations, set in official guidelines and assessments which cannot be ignored, often on pain of being reversed by courts.
Thus, although I am not aware of any empirical authority that is required to collectivise reason by voting according to the premisewise model, they are all under pressure, of various degrees, to collectivise reason. This pressure works above all informally, through the discourse in the public sphere. 11 Discourse works in the same way as voting in the abstract PWM model above: Any important decision will be preceded by public deliberation as to whether certain relevant premises obtain and the decision will be taken with some reference to the overall public opinion on these premises. There are various formal mechanisms that drive decision-making towards the premisewise mode too. The prime example is the separation of powers into different branches and departments, each of which is responsible for a subissue on which the ultimate decision is based. In the example above, the decision on P 3 may be in the hands of an independent Treasury, a designated Office for Budget Responsibility (OBR), or it may be reviewed by a court of auditors or congressional committee. Pettit (2004) extolls the virtue of a growing variety of contestatory institutions, where certain parts of the decisions-for example, conformity with constitutional requirements or human rights principles-can be challenged. In a growing number of cases, the decisions must be premised on the conclusions from regulatory impact assessments, opinions of expert advisers, or even decisions of international organisations. Many subdecisions are based on estimations or recommendations by the International Monetary Fund or the Organisation for Economic Co-operation and Development (OECD); issues of food safety are often preempted by the FAO/WHO Codex Alimentarius Commission, and the classifications by the European Chemicals Agency (ECHA) are reckoned with throughout the world. Sometimes even authoritative nongovernmental organisations like Human Rights Watch or Amnesty International may effectively determine the value of a pivotal premise for the decision of a sovereign government, e.g., did a certain authoritarian regime commit such atrocities that military intervention becomes morally justified. In many cases, decisions on certain factual premises may be taken just by copying the decisions already taken by regulators in other countries. If a certain substance is prohibited in a few neighbouring countries because it was found to be carcinogenic, the members in the decision-making panels will struggle to decide otherwise. Thus, even though collective decisions are not rationalised according to the model set out above, in my view all developed democracies do collectivise reason in a number of more or less informal ways. The effect of this collectivisation of reason is, of course, to make the authorities less responsive.
From this discussion it should be obvious what is the fundamental problem with elections (even in ideal conditions where all known pathologies and impossibilities are avoided). If we substitute the three individual decision-makers with three blocks of citizens of roughly equal size the dilemma remains. However, the incentives of elected representatives are strongly against collectivising reason. Even when the relevant premise-beliefs are identified (e.g., by opinion polls or focus groups), individual citizens will still make known their preferred policy conclusions, too, and the incentive for any elected representative to follow the latter rather than the former is overwhelming. Indeed, if elections are good for anything, it is to make decisions responsive, i.e., that they are taken in conclusionwise mode. However, if the argument of this paper is correct, responsiveness stands in the way of collective rationality. So, should we get rid of the responsiveness relationship and, if yes, should we discard elections too? In my view, the answer to both questions is still in the negative.
One may be forgiven for thinking that in a polity that collectivises reason the space of choice is severely constrained by the decisional matrix. But it would be wrong to conclude that choice is lost and collective reason amounts to the "TINA (There Is No Alternative) doctrine." It was already explained that the matrix is not produced by some external sages but is constructed within the political process and the set of premises is subject to constant contestation. Even when premises are taken for granted in a given moment of time, reasoning within the matrix can go in at least two directions. The more obvious one is from premises to the conclusion; however, panel members would often take their preferred conclusion as a point of departure and reflect recursively on their beliefs on the premises. 12 Further, new information may be brought forward tactically to support one conclusion rather than another. Thus, contrary to certain critics, in my view the rationality matrix does not depoliticize politics; it only makes it much more complex and shifts the choices from the overtly political (in the areas of the conclusions) to the more subtle choices in the area of the premises.

The Rationality Gap
Thus, notwithstanding its normative and practical appeal, collective rationality creates one obvious problem for democratic polities-too often it frustrates the individual wills of the many. 13 Certainly, appeals to reason to contain popular passion have been justified by a constellation of thinkers, e.g., Rousseau, Condorcet, Burke, Madison, and even Polybius. Indeed, almost all cannons of democracy seek to constrain the popular will for the sake of the common good, individual rights, principles of justice, etc. 14 Yet even while doing so, none of these thinkers abandons the notion that authoritative decisions are in some ways responsive to what citizens will. They could do that on the implicit assumption that by constraining factionalism or populist passions the "true" will of the people will reign. In one way or another, the collective will was equated with collective reason. Having recognised that there is a dilemma between the two, this is no longer plausible.
This inherent incompatibility of the will and the reason of the collectivity remains largely ignored in the literature. Indeed, thus far the discursive dilemma has been discussed only as a choice between two alternative decisional procedures. Actually, it shows something much deeper than this and bears on the very nature of political authority. If my argument is correct, the discursive dilemma is a choice between submitting to acts of collective reason (in extremis opposed by everyone) and adoption of arbitrary acts, which 12 I shall revisit this issue in Section 4. 13 The most notorious expressions of these frustrations are, of course, the vote for Brexit and the election of Donald Trump, which have commonly been described as revolt of the masses.
14 For a subtle account of this contradiction in American context see Bessette 1980. even if supported by overwhelming majorities, are not premised on reasons which are themselves supported by such majorities (i.e., for the majority of people there are no reasons at all). 15 As seen in the previous section, there are good reasons for the collective decision to be taken premisewise, despite the possible contrary opinion of the majority of the decision-makers or citizens. But if this is the case, can they be reasonably expected or morally required to suspend their own conclusion on the issue? Rousseau (and many others) answers in the positive, arguing that a citizen who goes against the majority is probably mistaken. 16 While there are good reasons to defer to the majority sometimes, 17 this cannot be accepted as a general rule, for it would be incompatible with the autonomy of persons as moral agents. 18 This would be a problem even in the case of individual members of a collective decision-making body, 19 let alone the case when the discursive dilemma arises for populations voting in blocs.
In the face of the discursive dilemma, Kornhauser and Sager rightly note that it is difficult to ignore the outcome desired by an overwhelming majority. They claim that collectivising reason "might be a democratically acceptable result in some circumstances, but it seems an improbable candidate for democratic sponsorship as a general rule" (Kornhauser and Sager 2004, 263). 20 Notwithstanding this, I am still in agreement with Pettit that if we cannot have both, collectivising reason is generally to be preferred as a normative ideal. More importantly, as explained above, every contemporary polity collectivises reason informally to some extent anyhow. But when we abandon the pursuit of responsiveness altogether, we are bound to face a considerable problem with the contrarian popular opinion which, is bound to persist.
This contradiction between what counts as collective reason and the aggregated individual reason I call the rationality gap. 21 It should be clear that the rationality gap is not a problem of the mere design of the institutions because the opposing majorities against many of the collective decisions will accrue spontaneously outside these institutions. In effect in any polity which collectivises reason both PWM and CWM, decision-making happen at the same time. When we opt to follow collective reason and gear the constitution accordingly, the contravening individual attitudes towards the conclusion will be formed simultaneously in the public sphere even if no conclusionwise decisions are taken anywhere at all.
In my view, here can be found the ultimate origin of populisms of all sorts, and their emergence is not a pathology of a democratic process gone astray but is its inevitable by-product. Various pundits explain the rise of populism with growing inequality, institutional incapacity, the complacency of the elites, cultural backlash, etc., 22 but on the suggested account these may only catalyse the structural problem. The discursive dilemma is certain to generate tensions between collectively rational decisions and popular opinion (i.e., aggregated individual conclusions) regardless of the substance of these decisions, and there always will be entrepreneurial politicians who identify these tensions to exploit them. If my argument is correct, whatever the policies adopted, it is likely that a rationality gap will open and create fertile grounds for populists. Hence they proliferate both in Scandinavian social democracies and in Anglo-American neoliberal states. In these circumstances, the problem is hardly the unresponsiveness of the elites. 23 If the elites are to blame for something, it is-as I shall explain in Section 4-for their failure to lead citizens towards collectively reasonable outcomes, i.e., for failing to close the rationality gap, and not for creating it in the first place. To sum up, the deliberative dilemma teaches us that to the extent that decisions are collectively reasonable, the unresponsiveness to "the people" is not a pathology, but the norm.
One may wonder why, if the rationality gap is so ubiquitous and persistent, our current frustrations have not risen long ago. I suspect the explanation is in the interplay of two factors. On the one hand, governance was not always that complex and issues so much interdependent; also, the institutions that drive decision-makers to PWM have not been so strong till very recently. Only since recently have governments had to take so many considerations into account, and only nowadays can these considerations be contested in so many institutional fora. While some pressure for collective reasoning has been always present, with the help of a number of internal branches and external organisations it has become by far stronger today than several decades ago. On the other hand, there are some factors that mitigate or obscure the gap which have grown weaker recently. One obvious factor is ideology-people with a shared ideology are likely to share beliefs on each premise-if A and B share a certain fairly deep worldview, they are likely to believe the same things on P 1 , P 2 , and P 3 . When this is the case the outcomes of the premisewise and conclusionwise procedure will be equal. With the demise of the traditional ideologies and party tribalism, it becomes more likely that their beliefs will be so distributed that rationality gaps will open on many issues. Thus, it might be the case that nowadays the collective will diverges from collective reason more often; hence, the frustration of the former has reached dangerous levels only now.
A final point to note is that the rationality gap may arise not only upon the adoption of new policies but also when the legitimacy of the established ones is challenged. While many policies are taken for granted by most of the public (e.g., austerity in Britain until very recently), none of them are insulated from new challenges by the critical public, so the leadership may struggle to justify them with acceptable reasons too. 24 Certainly, this is a difficult task, and yet it needs to be done on a daily basis. When the political leadership is called to justify such policies, they inevitably have to return to the basics on which these policies are premised and may face the discursive 22 For an empirical test of some of these explanations see Inglehart and Norris 2016. For the case of American populism, see also Rothwell and Diego-Rosell 2016. 23 The thesis for unresponsiveness of the elites as explanation of the origin of populism strikes me as an academic version of the populists' own rhetoric. 24 In the EU the free movement-of goods, services, capital, and people-has been enshrined in the Treaties since 1957, yet it is constantly under attack, both from the left and from the right, and neither the governments of the member states, nor the Union leaders have done a very good job of defending it. dilemma. Of course, pointing to the basic foundations of a controversial policy will not persuade its devout opponents, yet there is democratic value in the process itself, as it allows the citizens to think for themselves about the foundational premises of their own political preferences, and this might be the best antidote to the uncritical falling for the next populist. A plausible hypothesis is that our taking for granted of issues like universal human rights, equality, the rule of law, and so on allows us-and new generations even more so-to forget the implications they entail when deciding bread-and-butter issues like unemployment benefits, access to healthcare, fees for higher education, etc. 25

The Responsibility of the Elected Representative
Thus, in a condition of recurring tension between collective rationality and majority will, citizens are bound to live in a state of permanent frustration, which threatens the stability of the polity-it challenges its legitimacy, leads to apathy and cynicism, reduces choice at the ballot box, and breeds populisms which target "the system" as such. 26 Indeed, one's first intuition when facing the discursive dilemma is likely to be to throw the premise matrix away. But this may be difficult if the institutional framework is sufficiently strong and the public sphere sufficiently critical. All we can do is mitigate the adverse side effects. Thus, it is submitted that to be sustainable democracy needs a constant effort for alignment of collective reason with the majority will. Although Pettit seems all too ready to sacrifice responsiveness for the sake of collective reason and is not very much concerned about the individual frustrations, in his most comprehensive book on the subject (List and Pettit 2011), he and List suggest that the leadership of collective agents should be made responsible to seek a way out of the dilemma. They mean this for corporate management boards but the same reasoning applies also to state leadership.
In my view, governing elites in states are already responsibilized to close the rationality gap through elections. Arguably elections do not provide much of a meaningful input for democratic decision-making except on a few very salient issues. However, they are usually good in selecting certain individuals to "govern," 27 and this thrusts upon them the responsibility to mediate between the demands of collective reason and the urges of the popular will. 28 In other words, elections are the mech-25 Cf. Mair (2009), who also notes that these constrains are no longer self-evident for voters, and the parties in government need to justify them. Mair describes such issues as "depoliticized" but in my view the fact that a choice is constrained does not make it any less political; at most it may hide the politics. To the extent that this is so, we must seek to repoliticise the choice, i.e., to show the political consequences of apparently technical decisions on the premises. We should then either reestablish the political support for each of the premises or be ready to drop the commitment they support. 26 The defining feature of populism lies in its being a challenge to "the system" rather than to the incumbent government (Mudde and Kaltwasser 2014). 27 There are obvious exceptions of course, German elections in 2017 were followed by some six months of negotiations to form a government, and in Belgium and Italy such occurrences are quite common and much lengthier. 28 Actually, elections usually select parties rather than individuals to govern. The role of political parties in representation is well studied in the literature, but for my argument it is not very relevant whether the selected subject is a natural or legal person, so for the sake of simplicity here I refer to people only. anism that assigns the responsibility for closing the rationality gap. The precarity of the position of the elected representatives is key to discharging their function. On the one hand, they are responsible to adopt policies which consistently serve the collective purposes as enshrined in the constitution and which are justified by reasons acceptable to all. On the other hand, leaders must be responsive to electoral pressures if they wish to be re-elected. Thus, they are forced by the institutions of developed democracies to do the former and incentivised by elections to aim for the latter. They can accomplish both and keep their job only if they manage to close the rationality gap.
But if my argument that the dilemma is structural is correct, how can they even hope to achieve this? On a Schumpeterian account of democracy-where people vote for representatives who will transform electors' preferences into policy, this is not possible indeed. 29 However, the picture changes if we allow voters to change their attitudes after thinking about them for a while. As a matter of fact, citizens' attitudes are not constant and they are not exogenous for the political process, so if they are revised after deliberation in the public sphere the grim choice between frustration and populism might not arise. In the choice matrix in Section 2 the individual beliefs both on premises and on the conclusion were taken for granted and were aggregated by voting without prior deliberation. But this need not be so. For example, on P 1 , B and C may be able to persuade A that immigration is a real problem (or vice versa). Similarly, on P 3 , A and B may persuade C that the cost of the construction is not that high after all. Dryzek and List (2003) have already suggested that deliberation can help resolve some of the difficulties of the aggregation of individual beliefs or preferences into collective decisions and even claimed that deliberation is a necessary complement to any mechanisms for social choice. In that sense, the present proposal follows a well-established path. List and Puppe (2009) offer a comprehensive review of the impossibilities and paradoxes of judgement aggregation in particular and the contribution of deliberation to each. The common conclusion from this body of literature, as aptly put in List 2018, 647, is that "there does not exist a single, universally best aggregation rule. Choosing an aggregation rule requires trade-offs between different desiderata, and different solutions to those trade-offs are appropriate in different contexts." What sets this paper slightly apart is that the problem identified here is not in the aggregation rules-the premisewise model seems to deliver precisely what is expected. The rationality gap rather refers to an unintended side-effect when collectivised reasoning takes place in the real world. The solution I propose is inspired by another body of related literature, on deliberative polling. Fishkin (2011), Fishkin andLuskin (2005), and Farrar et al. (2010), to name but a few, provide empirical evidence that deliberation encourages reflection and people do revise their preferences. It is worth noting, however, that this literature assumes that the revised premises 29 Even though in my view this is a misguided understanding of democracy, it is remarkably common. In a recent attack against democracy tout court Jason Brennan explains that it sanctifies people's views such as they are. For criticism of the latter see Runciman 2018. It should be noted, however, that there may be some normative reasons to treat preferences as exogenous-in a liberal society citizens are free to believe and conclude as they please, and they cannot be held accountable either for their beliefs or for any errors in the inferences they draw from these beliefs. That is why any account which considers the possibility of such revision risks appearing paternalistic, manipulative, or worse. are ameliorated-they are more considerate, and the result of more informed judgements. As we shall see, however, this is not necessary for the job that is assigned to deliberation here-all that deliberation is expected to do is to lead to some alignment of the beliefs of A, B, and C, and some consistency in their decisions. While such decisions are expected to lead to less frustration, the judgements will not necessarily be better in any objective sense.
Such alignment may be made easier by the fact that reasoning does not need to go in one direction only. Agents may practice not only modus ponens but also modus tollens, so inferences may be drawn from the conclusions back to the "objective" premises. As these are both valid rules of inference, the representatives-in order to achieve a certain desired conclusion-may adjust their position on the premises accordingly. It is not irrational for someone, when confronted with inconsistency in their own beliefs and judgements, to update them by reasoning to and fro until they achieve a new equilibrium. Note that even when they do so in deliberation with others they remain as autonomous as ever, so the proposed account is by no means illiberal or paternalistic. Similarly, deliberating groups can go either way too. If a member is committed to a certain policy, i.e., to a certain conclusion, say, for constituency or coalition-policy reasons outside of the matrix, they can and will be likely to revise their own beliefs on the relevant premises in order to maintain consistency within (Goodin 2005). 30 More importantly, they will try to persuade their constituents into believing some facts which do support the policy. Thus, reasoning-individual as well as collective-will go in both directions: Beliefs on premises will determine conclusions, but also a commitment to certain conclusions may induce people to revise their beliefs on premises that stand in the way. While elected representatives are supposed to track both beliefs on premises and preferences for conclusions, they are in a position to mediate between the two. Thus, in the process of public deliberation, they ought to highlight that if someone is strongly committed to a certain C, this necessarily means that they accept P i to P j , and if this is not the case, they should revise either their initial beliefs or reconsider the commitment to C.
Contrary to the common impression, in my view, citizens are usually not irrational. 31 The reason public opinion often appears to be irrational is that citizens reason individually, and this has just been shown to be at odds with collective rationality. Those who still think the possibility of any deliberate, leadership-led shift in public opinion is utopian should look no further than the gradual movement of the British public from scepticism towards further European integration in the 1990s, to a vote to leave the EU in 2016, to the demand that this be done with "no deal" in 2019. 32 30 Of course, this may not always be possible. For example, wages may have been undercut by the inclusion of women in the labour force, but those fighting for higher wages would find it difficult to publicly position themselves against the inclusion of women. However, the inclusion of immigrants in the labour force does not represent similar taboo in the public imagination and remains an object of legitimate contestation. For this example I am grateful to Chakrabortty (2018). 31 It is true that people often hold views that appear inconsistent and therefore irrational. Cf. Tavernise and Cohn 2019, quoting Pew Research analysis according to which 54 percent of Americans hold a roughly equal mix of conservative and liberal positions. However, in my view this rather shows that there is ample space for these beliefs to change if the individuals who hold them are confronted in the right way. This, of course, would rarely be due to one speech, or one deliberative poll; it would rather be a gradual process over several years. 32 At the moment of writing, the "no deal" Brexit party is leading the polls for the forthcoming European Parliament elections.
Another example is the change of the Britons' attitudes to austerity. As Aditya Chakrabortty (2018) has pointed out, in two years Jeremy Corbyn's team turned what was once considered by everybody as an economic imperative into something that is now widely recognised as a failure. Similarly, Trump's climate change denialism has had a sizeable effect on the beliefs of Americans. 33 I submit that the reason for such shifts not happening more often is that far too many of us have internalised the Schumpeterian view of politics and have given up hope on trying to reason in the public sphere. Elected representatives are resigned to clinging on to the conclusions their constituents want to see rather than do their groundwork. But this need not be so: Communication in the public sphere-if it is to exist at all-is all about persuasion, which implies a revision of beliefs.
Thus, a discursive dilemma arises and persists only if representation is conceived merely as a transfer of information (about interests, preferences, identities, etc.) from the represented to the representatives. If we take representation to be a dynamic relationship, as Mansbridge (2003) 34 or Saward (2006) have it, it becomes possible for the representatives to reach a collectively rational decision and change the beliefs of the represented accordingly. Of course, they will do so only if they are forced to take collectively rational decisions and to justify them to their electorate in order to get reelected later. If Saward is right that the representatives construct new views of their constituencies, and their "interests," 35 they can be held responsible to square the circle and bridge the rationality gap. This is not to say that it is an easy task to do, 36 or that they always have the right incentives to do it. Certainly, in choosing how to make a representative claim, a rational leader would tend to prefer political "readymades" rather than go against the tide. But if I am right to believe that the institutions of contemporary constitutionalism drive decision-making towards premisewise mode, representatives are forced, to an extent, to collectivise reason. If they get reelected afterwards, this will mean that they have succeeded. 37 Thus, institutional failure may allow the government (as opposed to the contenders that are unconstrained) to turn to the popular will. On the other hand, strong institutions without an effective re-election cannot tolerate incumbents who act according to collective reason but who do their deliberative job poorly and fail to close the rationality gaps. Thus, the 33 In 2018, 65 percent of independent voters were aware that there is scientific consensus that global warming is happening; this is 6 percent less than in 2017 (Brenan and Saad 2018). 34 In her anticipatory mode of representation Mansbridge emphasises the focus on the "projection"-of where the representative expects the represented to stand in few years' time, but also where she expects to bring them herself. She also sees the key role of public communication, commenting that "the better the communication between voter and representative [...], the better the representation" (Mansbridge 2003, 519). Similarly, Urbinati emphasises the importance of "circulation of judgement and influence between institutions and society [that] has political parties or movements as its indispensable pivoting forces" (Urbinati 2011, 46). 35 Saward (2006, 303) notes that representatives are not free to do that as they wish; the claims that are compelling for the relevant constituencies must rely on existing terms and understandings. 36 There are examples from the worst of circumstances-of postapartheid South Africa and former Yugoslavia, where divisions on fundamental values were overcome by responsible leaders (Vasilev 2012). 37 Note that on the dynamic understanding of representation the question whether elections provide adequate information for the beliefs and preferences of the represented becomes somewhat moot. Still, those who get elected are presumably the best experts in both tracking people's beliefs, and changing them. EU suffers from greater democratic deficit than its member states because Mr. Juncker is not forced to justify the Commission's policies to the European citizens and go to the polls thereafter. 38 The precarious position of the representative is famously illustrated by the political misfortune of Edmund Burke. He was rightly convinced that as a responsible member of Parliament he should support policies that would further collective goals (i.e., what is collectively rational in my terms). Yet in order to get re-elected, he also had to persuade his constituency that his vote was in their common interest properly conceived. The first duty Burke fulfilled, voting against an import tax his constituents wanted; however, he failed to achieve the second part and accordingly lost his office in the next elections. 39 In our own times, when the convergence of the policies of mainstream parties towards the centre is much lamented (this convergence may well be a consequence of the rationality gap rather than its cause) and people do not find many alternatives on their ballots, politicians are left without meaningful occupation. 40 Therefore, they may be assigned the responsibility to lead public opinion in such a way that the gap is closed.
I would like to stress once again that the word persuasion, which has an unavoidable air of paternalism, in this context denotes a process of reasoning together, which leads to alignment of beliefs and conclusions but not necessarily in the direction desired by the leader. Therefore, it should not be understood in the Aristotelian sense of a governing elite telling the common citizens what is best for them. The careful reader should remember that collective reason is defined on the basis of the actual beliefs (but not policy preferences) of "the people." Thus, it does not presuppose any standard for individual rationality. 41 Leaders, on this understanding, are not people who know something better-often the opposite is true. They may, however, have a better ability to track people's actual beliefs and better recognition of the constraints of the matrix if they can win and remain in office.
The decisional matrix does constrain choices but the matrix itself is constructed by the relevant citizens. 42 The experts may have had stronger voices on one premise 38 Sylvie Goulard, MEP, speaking in 2014, aptly noted that Coca-Cola launches a huge campaign when introducing a new cola but no one bothers to do so when introducing a major change in EU policies. Paradoxically, Eurosceptics may do a better job bringing the EU closer to people when they contest certain policies, for it is only then that the leaders of the EU are forced to justify them in the public sphere. 39 For Burke the role of the elected representative was to be an epistemic authority, owing his judgement to the electorate, and this has some air of elitism. On the suggested account the representative is not that (or not only that), but an expert in negotiation and mediation, an expert in opinion formation, i.e., in gap-closing. Thus, experts might have to pay closer attention to the concerns of their constituents and might have to revise their own beliefs. 40 For example, Mouffe (2000, 72) claims that the boom of right-wing populists is because they are the only ones that offer an alternative that citizens can vote for (even if this alternative is unworkable and morally abhorrent). 41 I fully agree with Bickerton and Accetti (2017, 186), who note that "once we accept the assumption that there exists an 'objective' and pre-political standard of political legitimacy, it seems difficult to resist the conclusion that some individuals are more competent than others in knowing what this standard should be and therefore be entrusted with the responsibility for ruling others." 42 It should be obvious that only some premises relevant for any given decision will gain wide recognition as such. There will always be many other factors for the decision that for many reasons will remain marginalised. Further, there is an obvious trade-off between a collective decision that is inclusive in that sense, and understandable governance where one or two factors carry the day. For reasons of space these issues will be explored in my future work. or another but their voices are only ones amongst many in the public sphere and in principle no one has any privileged access to the drawing board. Decision-makers (A, B, and C above, or the segments of the population they represent) can and do choose what to believe and may be strategically guided by their preference for a given conclusion. Yet the choice is essentially collective, so much so that often it will be difficult to attribute it to any leader or, as we have seen, to any particular majority. What the leaders can and ought to do in these circumstances is to make the matrix-that is, the imperatives of collective rationality-clear to everyone. They have to publicly show that a given popular belief warrants a certain course of action and bars another, so that if someone remains committed to the latter, they must give up their initial belief. The leader ought to expose only the relationship between popular premise beliefs, prior commitments, and policy choices. Surely, they also advocate for one course of action and against another, but their duty is to act according to collective rationality, which remains firmly anchored in the public beliefs such as they are in the relevant moment. What a responsible leader ought not to do is to advocate for actions which are not compatible with these beliefs or prior commitments without bringing about their revision. In other words, they should not claim that we can have our cake and eat it too. 43 The task of responsible leaders is to first identify and then justify the course of action which is supported by the relevant beliefs of the public. The justification is a matter of logic, epistemology, and, of course, rhetoric-it takes constructing a coherent story, based on believable assumptions and undefeated facts. What the actual beliefs of the people are is an empirical matter, and whether they support the proposed course of action is a matter of degree. These are generally knowable and measurable-by opinion polls, focus groups, etc. Despite appearance to the contrary, these beliefs do not always determine the conclusion fully, and as at least some of them are potentially subject to reversals, the decisional matrices can sustain more than one conclusion. Thus, political leaders offer competing bundles of policies that are justified with certain reasons, often shared between different policies in the bundle. These bundles may win public support, and if they do, some citizens may revise some of their own beliefs accordingly. If the contestatory institutions work as they should, only policies that are internally coherent and consistent with the previous commitments will find their way to the statute book. If these policies are justified by reasons that are believed by the citizens, their proponents are likely to win the next elections. As Damian Chalmers (2000, 190) puts it, the representative's role is to be the "storyteller-in-chief." Deliberator-in-chief would be more precise but a bit too clumsy for a job title.

Degeneration of Collective Reasoning
In the previous section, I discussed the role of discourse in the public sphere for the closing of what I called the rationality gap. Actually, it does more than this-public discourse is essential to maintaining the relationship between the premises and the 43 The infamous insistence of the current British prime minister that after Brexit Britain can retain full access to the single market, and regain control over immigration, is a paradigmatic example of failure of leadership (Dunn 2016). conclusion. 44 As was suggested, the most common method for collectivisation of reason is by dispersal of power in such a way that the decisions on each of the premises (P 1 to P i ) is allocated to a dedicated institution which is made responsible for tracking the relevant beliefs-of the general public, or of the relevant epistemic community: economists, physicians, etc.-and issuing an authoritative statement on this particular premise. From the perspective of formal logic, this seems a good approximation of the collective reasoning model-each premise is determined independently, with no regard to the other premises or the anticipated conclusion. Yet there is a danger that this would remove the collective from collective reasoning. The careful reader will remember that in the abstract model defined above, all panel members take part in the decision on each of the premises. Only in this way can the premisewise decision be considered collectively rational, i.e., based on beliefs that are shared by the members of the collective. This is not the same as a system where each of the members, A, B, and C, is made exclusively responsible for the determination of one premise each. While the latter may be useful as a power-sharing model, these decisions will not be collectively rational.
Arguably the latter is often happening in contemporary polities and may be described as technocracy. 45 Assigning the decisions on particular premises to specialised institutions is one of the steps necessary to attain collective rationality; however, the judgements of these institutions shall not belong to them exclusively. If these judgements are to stand for the collective beliefs on the respective issue, any such judgement must be communicated to the rest of the public and be effectively shared by this public. This may be easier than it appears, as most of the public will not have strong feelings on most of the policy-relevant premises. Still, there always will be some members of the public-or specialised civic organisations-who are ready to challenge the view of the respective institution. In distinction to technocracy, in a system for collective reasoning, the specialised institutions remain open to such challenges and do not enjoy unconditional deference by the other decision-makers. Which way the final decision and the public will go depends on the general trust any given institution enjoys and the deliberative effort it has deployed in the particular case.
A common objection here is that there are some "technical" issues that can and even ought to be isolated and delegated to specialised bodies. Yet it is submitted 44 This statement is epistemologically controversial. One could argue that the relationship either exits or it does not, and this is independent of whether there is anyone to talk about it. However, as a practical matter, if a certain premise is not subject to intense public debate, it is not very likely to be taken into consideration by the decision-maker. 45 Technocracy is usually defined as "rule by experts" and is based on the assumption that certain problems have efficient solutions which may be known in advance (Radaelli 1999, chap. 2). Neither of the two is true in a system that collectivizes reason in the way suggested here. In our matrix A, B, and C have equal epistemic authority for each of the premises, and even after deliberation they may defer to each other only if they have revised their views voluntarily. Also, collectively rational decisions are normatively preferable not because they are superior but only because they accord with the relevant premise-beliefs and in this way they will be consistent with the other decisions of the relevant group. The only commonalty between collective reasoning and technocracy is that in both cases decisions are not responsive to the preferences of the majority. Yet the two systems respond to the latter problem differently. While a technocrat would want the public to be "educated" to accept the optimal decision, in the preceding section I argued that reasoning may go either way. that rarely can an issue be considered as purely "expert" and Science, Technology, and Society (STS) scholarship has abundantly shown that even most technical issues involve judgment (Jasanoff 2004;Paskalev 2019) and that these judgments are unavoidably subjective. They are profoundly affected by the context they arise in and by a number of implicit assumptions decision-makers make. Accordingly, when the issues (premises) are decomposed and allocated to different "black boxes" the influence of certain implicit values and worldviews is hidden and the relevant contexts are ignored. To use again the example in Section 2 above, on P 1 , even if we assume that the adverse effects of immigration on the job market are fully known and certain, the question whether this is a problem for the society as a whole remains contested, and obviously depends on which segment of society is concerned (and it may also vary with the differences in immigration too). Further, beliefs on the premises may be subtly interrelated, even if the premises concern relatively independent issues. For example, the cost of a certain policy (P 3 )-say 1 percent of GDP-may be considered unfeasible in the context of unemployment, but feasible in the context of national security (or vice versa). While a polity which seeks to collectivise reason may very well employ an independent financial authority to consider this question, this authority should only inform the other decision-makers of the budgetary implications and possible trade-offs, but every effort should be made to prevent this advice from amounting to an effective veto (i.e., for A to dictate P 3 to B and C) which preempts the overall decision. The latter belongs to the collective as a whole, with each member deciding by taking everything into consideration; this cannot be relegated to any part thereof.
When a specialised body is made solely responsible for determining one of the premises (de jure or de facto), it is implicitly assumed that the other participants have no opinion on the matter, which will only sometimes be the case. In such cases, our model for collectivized reason degenerates into something different-technocracy or worse-because the decisions yielded will no longer be formed on the basis of publicly supported beliefs. On the contrary, such decisions would render the individual beliefs irrelevant or at least discounted. When the judgement of A on P 3 is allowed to carry the day-say, because of A's superior expertise on P 3 -the beliefs of B and C on P 3 are treated as if they were zeros. In some cases, B and C may adjust their own beliefs because they trust A's superior knowledge, because they were given persuasive arguments, or were shown some contradiction with their other beliefs. However, it cannot be assumed that in the general case citizens will suspend their own beliefs as soon a specialized institution has issued its opinion.
Thus, while any given policy decision may need to be decomposed into a number of subdecisions in order to ensure that the policy is justified by publicly supported reasons, the assignment of the subdecisions to separate institutions is tricky. On the one hand, this may be the only way to ensure that the separate subdecisions are taken seriously. But on the other hand, there is a risk that the institutions are too autonomous and become dictators on "their" issue. For really collectivized reasoning, any such institutions should aim to inform the general discourse without fragmenting it. Decomposition without continuous discourse represents a degeneration of collective reason and it can take two forms. Collective reason will degenerate mildly when people trust the dictator on a certain issue, so much so that they do suspend their own judgement. In principle, trust is good and allows a certain economization of reasoning without which life in society would be impossible. However, beyond a certain point, it may lead to a decline of moral autonomy and civic virtue. Ideally, trust should be complemented with contestations: vibrant NGOs specialized on the issue which regularly challenge the authoritative pronouncements so that citizens are at least potentially able to form informed opinions on their own. Note that on this account contestation, which stands for reason-giving, persuasion, and general criticism in the public sphere, is an essential component of the system for collective reasoning, not (only) a mechanism for correction of errors. In the severe form of degeneration, people do have opinions, but they are substituted by the expert authority, which dismisses the contestations as uninformed or irrelevant. Institutions turn from fora to veto-gates, and instead of public discourse we have turf wars.
In this section, I have strived to distinguish collectivised reason from technocracy. Indeed, there are similarities in the way each of the two is implemented in practice. Both systems appear to be "depoliticised," and rely on the involvement of independent agencies comprised of experts, and in both cases the decision-makers' choices are restricted by a dense web of rules and requirements for justification. Yet in the technocratic case, the presumed superior knowledge is recognised as a source of authority in its own right, which precludes further debate, while in the case of collectivised reasoning every reason and every subdecision is potentially open to challenge. Specialised institutions may have only imperfect authority over the relevant premises in order not to preempt external challenges and to maintain healthy and meaningful public discourse. 46 If they were to have complete power, their opinions would follow exclusively the internal dynamics of the institution and would have no need to engage in meaningful discourse with any other agents. On the other hand, if the specialized institutions have no power at all, they may be sidelined by those who have power and in turn by any other rational actor. Ideally, specialized institutions should be focal points of the discourse on certain issues and part of a dense web of agents with overlapping competencies and worldviews that carry the discourse.

Populism and Leadership
The previous section discussed one possible failure of collective reasoning, that is, when the institutional framework which is to drive decision-making into premisewise mode is too rigid and the relevant discourse in the public sphere is stymied. Now we turn to the other end of the spectrum, where the framework is too weak, so that it allows the decision-makers to abandon the pursuit of collective reason and resort to the conclusionwise option. As was already suggested the rationality gap is the ur-source of populism, and indeed, according to Mudde, one of the defining features of populists is their willingness to bypass any institutions which are at odds with the popular will (Mudde and Kaltwasser 2014). 47 While for many critics 46 In global administrative law Krisch (2017) calls this "liquid authority." For example, the decisions of the International Agency for Research on Cancer (IARC) are not binding on anyone, yet they are taken into consideration in every decision for which they are relevant (Paskalev 2019). 47 Populists also peddle their crippled epistemology to borrow Hardin's (2002) phrase, which omits some or even most of the premises relevant for policy choices. Thus, they undermine the rigour and sophistication of public discourse impoverishing collective reason and barring any possibility for closure of the rationality gap. The importance of the completeness of the set of relevant premises and the general quality of the public discourse go beyond the scope of this paper. the "will of the people" is merely posited by the populists, on the suggested account it is all too real and represents the aggregation of the conclusions in our matrix.
In this section, I will take Mudde's definition as a point of departure and will consider how the phenomenon identified in Section 3 above can contribute to the explanation of populism. In contrast to the populists, who exploit rationality gaps, responsible leaders can be defined as those who side with collective reason and who struggle to close the rationality gap by deliberation in the public sphere as suggested in Section 4 above. The difference is between those who take public attitudes for granted and those who seek to change them, whatever the direction of change. This may appear as an odd candidate to be assigned normative importance, but it reflects the need for enormous deliberative effort for aligning beliefs, an effort without which the rationality gap cannot be closed. Thus, leaders can be defined as those who invite reflection and revision of beliefs, while populists are those who reinforce these beliefs. In the remainder of this section, I will only begin to apply the suggested conceptual apparatus to one salient example where populism is believed to be at play. For reasons of space, a more comprehensive analysis of concrete empirical cases will have to wait for future work.
People in positions of power are subjected to contradictory incentives. Obviously, they always have the incentive to follow the aggregated will of the majority rather than to struggle to align it with collective reason; it is by far easier to be led rather than to lead. On the other hand, as we saw, this temptation is tempered-more or less effectively-by strong contestatory institutions. The former incentive, however, is often strengthened by the mere availability of contenders who are not restrained by the responsibilities of office and are free to be fully responsive to the majority opinions that aggregate spontaneously in the public sphere. Thus, to take but one example, the virulent anti-immigration rhetoric of the UK Independence Party (UKIP) gradually infected both Conservatives and Labour and brought about momentous changes in the relevant policies of the country without UKIP ever been elected to office. Conversely, once in office populists sometimes move toward the mainstream as they are subject to the said institutional constraints. 48 On the suggested account the current US president, Donald Trump, is a classic example of a populist. Mr. Trump does not appear to have any consistent set of beliefs, and on any policy question that becomes salient, he seems to be just embracing the solution that the majority of American citizens seem to prefer. He cannot be bothered to reflect on any considerations these policies are premised on, let alone to invite reflection by his constituents. Most importantly, his policies may make a U-turn from one tweet to another. In contrast, on the suggested account the former leader of UKIP and current owner (sic!) of the Brexit Party, Nigel Farage, appears to be a normal leader. Mr. Farage has strong views on a certain policy issue, so he joined a few likeminded fellow citizens in launching a political project some twenty years ago and 48 But see Cas Mudde (2019), who notes that the recent success of Poland's Law and Justice party calls into question the "myth" that populists moderate in office. One possible answer is that it is precisely because populists come under the pressure to collectivise reason that when they are strong enough they launch sustained efforts to undermine the institutions that enforce it in the name of "the people," e.g., the majority will.
persisted over a number of elections (and electoral failures!). No matter how silly I personally find his views, they are remarkably consistent on their own terms. Most importantly, over the years he managed to push his pet issue from the conservative fringes to the centre of public discourse, and eventually brought a large part of the society along with him. When UKIP was set up in the 1990s most people did not share all of his beliefs on the premises relevant for the decision to leave the EU. Yet as the years passed, they seem to have updated them to fit the conclusion preferred by Mr. Farage. 49 To see this we may employ our matrix again. Certainly, there were many more than three premises that were relevant for the decision for Britain to leave the EU and it is not my aim to reconstruct the actual path through which the collective decision was arrived at. A highly stylised version will suffice to show how the analysis could work. Thus, we may consider a matrix with the following three premises: P 1 -whether immigration from the EU is a major problem; P 2 -whether leaving the single market will hit the economy too hard; and P 3 -whether the European "dictate" is stifling national sovereignty. It is beyond the scope of this paper to trace the fluctuations in the actual beliefs of the British people on these issues, but it is plausible to assume that in 2016 there was a majority believing that P 1 and P 3 obtained. Indeed, all major parties had manifesto pledges to "control" immigration, and the then prime minister, David Cameron, had explicitly made the reform of the free movement rules pivotal for his own position. His failure to achieve any substantial change of these rules in February 2016 could not help but reinforce the beliefs that immigration is a problem and that it cannot be managed within the Union, so his subsequent campaigning for Remain was doomed on this count alone. Beliefs on P 3 were less clear, but there was no major party making a positive case that EU regulations do any good for Britain. While the Leavers could ride the wave of the familiar arguments about the free market, there were very few voices making the point that these regulations protect workers, consumers, public health, and the environment. The strongest statements the Remain side made was that this is the cost to be paid for the benefits of the single market. This, of course, implies that the regulations from Brussels are a bad thing in the first place. Thus, with most public voices admitting (and probably most of the citizens believing) that P 1 and P 3 obtain, the premise that becomes pivotal is P 2 . If a majority of people did believe that there would be no economic downturn (or that it will not be significant, or that the importance of this issue is not on a par with the others), then the decision to Leave would have been collectively rational. 50 This conclusion may appear surprising as it goes against the nearly complete academic consensus that Brexit was a high point of the populist tide in 49 As is well known, Britain was almost the only one of the "old" member states that did not impose any restrictions to free movement of people from the Eastern European countries which joined the Union in 2004. Now it is difficult to believe that this was a matter of course only some fifteen years ago! 50 Obviously, this example is deliberately simplified to illustrate how collective reasoning works. For an analysis of the real reasons for the referendum result, a much larger and more nuanced set of relevant premises needs to be identified with, of course, empirical data about the actual beliefs of people on these premises.
Europe. 51 But this surprise should only bring home the point how very different collective rationality may be from what is rational for an individual observer. On the terms of this article, collective rationality is defined solely on the basis of the relevant beliefs of the decision-makers and, in this case, of the population as a whole. Whether these beliefs are "objectively" correct is an entirely different question. I hope the surprising conclusion will also dispel the air of elitism that words like leaders, reason, persuasion, and deliberation inevitably carry. While my argument is that responsible leaders ought to encourage reflection on the policy-relevant facts and seek to establish the truth, the one thing that determines the outcome of collective reasoning is the actual beliefs of the citizens at the relevant moment. Thus, collective rationality does not claim any epistemic superiority. Our reasons to prefer premisewise to conclusionwise decisions were set out in Section 2 and they are grounded in legitimacy, nondomination, and agency but not on epistemology. On the contrary, I find nothing to suggest that any of the two modes is better than the other on the latter ground. It should be noted, however, that public contestations and the deliberation that is called for to close the gaps can be expected to intensify the connection with "reality." In the same vein, the careful reader may wish to note the volatility of public beliefs on the premises and question the legitimacy of the decisions based on them. Indeed, public beliefs are susceptible to targeted disinformation campaigns or may reflect the established power relations. But again, there is no reason to expect that the premisewise mode is any more or any less vulnerable to power and manipulation than conclusionwise mode. With both modes there is the problem, absent any internal yardstick, of judging the quality of the outcome (any external yardsticks would be unacceptably elitist). So if we want to remain democratic, decisions must be based in one way or another on the relevant popular attitudes, where preferences (on the conclusions) and beliefs (on the premises) are (un)reliable in equal measure. Even if "we know" that some popular beliefs are patently false (anti-vaxxers come to mind) they should not be dismissed as irrelevant or rubbish, and this is so not only for reasons of democracy. In my view, there are good epistemic reasons for false beliefs to be taken seriously, as they may be a symptom of something that is real (e.g., failing trust in public authorities in the case of anti-vaxxers). Thus, the concerns of the people-real or perceived-should not be discounted, they must be the point of departure for any policy discussion. If we as academics wish to ameliorate the epistemic basis of democratic decision-making we cannot do anything better than get involved in the process, as citizens. 52

Conclusion
Drawing on the existing literature in this article I discussed the distinction between collective reason and collective will. I identified a problem neglected in this literature-the systemic frustration of public will by collective reason that I called the rationality gap. While elected representatives have every incentive to be responsive to the aggregated preferences of citizens, I argued that in the developed democracies of today there are a number of contestatory institutions that do not allow them to easily do so. These institutions, and more informally critical discourse in the public sphere, force them to decide within the bounds of collective reason. This is the stick meant to assure that the leaders are responsible rather than responsive. On the other hand, elections provide the carrot to make sure they do the deliberative work as outlined in Section 4 and bring collective reason and the collective will together. The dual pressure of contestatory institutions and elections defines the role of political leadership to do its best to close the rationality gap.
I believe this analysis is helpful for understanding the root cause of populism, which occurs when the institutions-formal and informal-meant to sustain collective reason fail to do so. More importantly, even when they succeed, they breed their own protest: Successful collectivised reasoning is bound to open rationality gaps. Thus, populism is not contingent on certain design flaws or policy failures; on the contrary, it appears when institutions do sustain collective reason but the incumbent decision-makers fail to adequately address the rationality gap. If we are to draw a policy prescription to respond to the current wave of populism, it is that more institutional checks are needed to safeguard the capacity for collective reasoning. The informal pressure exerted by discourse in the public sphere is increasingly insufficient and needs enforcement through more contestatory institutions. This is hardly new. However, if my argument is correct, it should be clear that this alone will not be enough. Even if it succeeds in safeguarding collective reason, the institutional buttressing of collective reason should be matched with a heroic increase in the deliberative efforts for closure of the rationality gap. Unfortunately, the response so far seems to have been in the other direction, with mainstream parties mimicking the populists and embracing their agenda. Perhaps the most consequential example here is David Cameron's pledge to reduce immigration from 300,000 a year to "tens of thousands," which was clearly incompatible with the other commitments of the country and eventually lead to a revision of one of these commitments.
It bears one more repetition that the role which leadership, so defined, plays in the very functioning of democracy does not imply any elitism or paternalism. Our representatives lead only in the sense that they are elected to office for their competency to track popular beliefs, reason in the public sphere, and help the formation of public opinions that are collectively rational. They are supposed to have a special competence for bridging the rationality gap, but both their point of departure and their endpoint reside with the people.
I believe this article has made two further contributions. The first one is normative-to assign responsibility for the closure of the rationality gap to the relevant decision-makers. Although this is a mammoth task, it is hardly impossible, and evidence from deliberative polls suggest that it works. This prompts one even more far-reaching normative aim-for the future work of the present author and, hopefully, readers-which is to substitute the predominant "Schumpeterian" paradigm in politics and in political science which takes people's preferences and beliefs for granted and leaves politicians responsible only to track them, with a new one, where beliefs are endogenous to the political process and leaders are expected to change them. The second complementary contribution is analytical. As was briefly sketched in the last section, the model for collective reason can be used to analyse a salient example of alleged populism, with sometimes unexpected results. More generally, it follows that populism is a systemic rather than a contingent problem.