The Epistemic Value of Democratic Deliberation

The Epistemic Value of Democratic Deliberation For article in Oxford Handbook of Deliberative Democracy Word count: ~7,800 David Estlund Helene Landemore DRAFT (for editing before submission) January 13, 2016 I. Introduction Deliberation, roughly the weighing of reasons, is something individuals do for themselves, but it also has an interpersonal (or “intersubjective”) form--the collective weighing of reasons with others, by communicating, arguing, debating, and persuading. Democratic deliberation is not only interpersonal, but also public and structured in ways necessary to count as democratic, a matter handled differently by different theories of democracy. 1 Deliberation has long been valued by deliberative democrats for reasons that have to do more with its intrinsic properties or the by-products it generates rather than what some now see (and some saw all along) as its primary point: figuring out the truth. Probably influenced by John Rawls’ famous stance of “epistemic abstinence” (Rawls 1985, as read by Raz 1990), early deliberative democrats mostly focused on the expression of respect and equality 1 that letting everyone speak and exchange reasons for their views before deciding on them was supposed to represent. Others emphasized the airing of grievances, the mutual understanding, the consensus and community-building that deliberating together was taken to allow for. Only recently have so-called “epistemic democrats” been paying attention to the more purely instrumental value of deliberation: maximizing the chances of getting to the correct or right decision, or at least getting as close to it as possible. By “correct or right decision” here, or “the truth,” can be meant an array of things, from objective truth of the matter (about facts or morality) to a more intersubjective, culturally-dependent, and temporary construct (about more socially constructed facts or moral questions). What epistemic democrats emphasize, on some readings, is merely the Habermasian (and commonsensical enough) point that we wouldn’t be exchanging reasons in the first place if we did not believe that there was something to figure out, whether we call this something the truth, the right, or the correct, just, or socially useful answer (see also Marti 2006). Epistemic democracy was first articulated in a 1986 article by Joshua Cohen entitled “An epistemic conception of democracy,” which was borrowing from Jules Coleman and John Ferejohn’s epistemic interpretation of voting. Cohen’s article was written in an attempt to defend so-called populist approaches to democracy (such as Rousseau2 and on some readings John Stuart Mill’s) from 2 the liberal conception of Schumpeter and Riker, which supposedly did away with the problematic notion of a “common good” or “general will” notoriously found in populist authors. Cohen proposed an interpretation of the populist view meant to be both coherent and more plausible than Riker’s own interpretation. The main move consisted in showing that Riker’s attribution of a “pure proceduralist” position to populist democrats was a mistake. He instead identified their view as epistemic. Cohen then goes on to characterize an epistemic interpretation of voting as having three main elements: 1) an independent standard of correct decisions 2) a cognitive account of voting (voting is supposed to express views about what the correct policies are according to the independent standard) 3) an account of how people adjust their beliefs in light of other people’s beliefs Cohen then proceeded to plug this interpretation of voting into a more general epistemic conception of democracy, one in which one finds all three elements 3 above applied to not just voting but other forms of democratic decisionmaking as well, such as deliberation. Epistemic democracy was further developed as a distinctive approach to democracy in the work of David Estlund (1997; 2008). Revisiting the question of political authority, Estlund argued that pure proceduralists (whether deliberative or aggregative) were mistaken to believe they could ignore questions of epistemic competence and performance. He offered instead a new philosophical framework—epistemic proceduralism—to reconcile concerns for procedures and outcomes. Estlund further argued that the minimal epistemic performance one should expect from a political authority must be set at ‘better than random’ and conjectured that democracies met that threshold, though probably underperformed compared to what he labeled “epistocracies” (regimes in which the few “knowers” rule). One epistemic engine he considered obvious was deliberation. Since then, a number of other epistemic democrats have contributed to the paradigm and pursued the complex philosophical questions opened up by an epistemic approach.3 When it comes to the question of epistemic (democratic) deliberation per se, two dimensions—normative and descriptive/explicative— have been explored. The normative dimension covers questions such as: Does democratic deliberation need to have epistemic properties in order for its outcomes to be normatively authoritative/have legitimacy? Or is public exchange 4 of reasons among free and equal valuable in and of itself, even if it tends to yield, say, more polarized views? How should we balance epistemic and other properties of democratic deliberation in an all-considered evaluation? The descriptive/explicative dimension includes questions such as: Does deliberation have epistemic properties? Does democratic deliberation have distinct epistemic properties? If so, could those explain the success of democracies in the real world? What role does consensus/unanimity play in these properties? Should democratic deliberation aim at consensus? What is the relation of democratic deliberation to majority rule? The rest of this article is organized as follows. Section II takes stock of the debates in the field of deliberative democracy that have been generated by advocates of the new paradigm of epistemic democracy. Section III turns to the epistemic properties of deliberation per se and, specifically, democratic deliberation. II. Normative frameworks A. Epistemic democracy: why? There are two distinct things that might be signified by speaking of an epistemic element in a democratic theory. One thing that is often meant is that the right or best decision is produced by voters thinking and reasoning about what would be right. This element is clear in Cohen’s account, described above, and 5 such accounts are usefully distinguished from accounts in which the decisions tend to be good or best, but by virtue of some invisible hand mechanism, for example sectarian interests cancelling each other out in the aggregate. A second and distinct meaning for an epistemic element is to emphasize the discovery of truths rather than merely producing valuable results. Indeed, as we will see (see description of Landemore 2013 below) some approaches emphasize mechanisms that are held to be truth-revealing quite apart from whether the truths in question are also morally valuable—which, of course, they are also held to be in such accounts. It is not hard to see why an epistemic element might be important to a theory of justified (or legitimate or authoritative) political rule. Otherwise, why not have rule by the person or procedure who can best determine what ought to be done? Traditionally, this question is a challenge to democratic thought since elites or experts can seem likely to outperform the body of all citizens consulted together. Maybe no specially chosen subset whatsoever could do better than democratic arrangements, but that’s a strong and contestable claim. Some advocates of democracy eschew any epistemic criteria at all, relying on features of democracy that are meant to justify or require it entirely apart from any claims about the substantive quality of its likely decisions. Public deliberation, however, is appealed to by many as a potentially powerful epistemic engine, in which case the epistemic and the democratic strands of our thinking 6 might not be in conflict. It would be possible, of course, to argue that democracy might tend toward correct answers on some question, and even by way of interpersonal deliberation, even though voters are not seen as addressing that question. This counts as a kind of invisible hand case, and there are such mechanisms in some contexts outside democracy such as certain idealized economic markets. On this approach some plausible mechanism would need to be described for the democratic context. Normative democratic theory that insists on an epistemic element does not necessarily make any claim about whether democracy (actually existing, or even in idealized forms) has epistemic value. Rather, it claims that unless democracy does have epistemic value there is no adequate case for its legitimacy, or authority, or justification (whereas if it does there is). So the interest in epistemic approaches can arise initially out of theoretical difficulties faced by thoroughly non-epistemic alternatives. That would tempt one to wonder and explore whether introducing an epistemic element can avoid the problems. Then, of course, the further question arises as to how appealing to (the right kind of) epistemic value would solve those problems without raising worse ones. For example, one of the more prominent worries about epistemic approaches is that whatever reasons there are for introducing an epistemic element would 7 also exert theoretical pressure to abandon democracy (partially or altogether) in favor of rule by some wise subset of subjects.1 There are non-epistemic but still instrumental approaches, a category mentioned above (Riker 1982, criticized by Cohen 1986). A more popular route is to eschew all appeal to instrumental value of democracy including epistemic versions of it, in favor of some value deriving wholly from a decision’s origins in a democratic process. An influential line of thought is that voting is a procedure that is fair to all who have the right to participate. A decision made by, for example, majority rule might count as fair in that way quite apart from whether it is a good decision on procedure-independent grounds.2 For example, it might be an unwise decision or even unjust, but, in its favor, it is at least the outcome of a fair procedure. If the measure of fairness is each participant’s having an equal chance of being decisive, then this would not recommend a voting procedure over choosing the outcome by a random procedure such as a coin flip. That procedure, like majority rule, gives no one more power in the procedure than anyone else. The value of procedural fairness of that kind is not obviously enough to call for democracy. Such considerations might suggest that a plausible theory will indeed give some role to a tendency to make substantively good decisions—at least, say, better than random. 1 Jason Brennan (2016) follows this line of argument to what he claims is, indeed, its antidemocratic conclusion . A crucial question for epistemic deliberative democracy is how that slope might be made less slippery. 2 One important account of the value of democracy in terms of procedural fairness, for example, is Waldron (1999). 8 A somewhat different approach from that of procedural fairness—still noninstrumental and so non-epistemic—is to argue that individuals have an equal moral right to participate equally in political decisions as an implication of a requirement of social justice whereby all citizens are of one class or status and none is inferior (Christiano 1996; Pettit 2012). Of course, if everyone is equally morally important or deserving, then it would seem to be of high importance that political decisions give everyone what is due to them as a matter of substantively just decisions, and not just in the decision procedure. The equal status account would need to explain why the equal status of having a vote is so morally weighty that it should be respected even if doing so would, by eschewing more epistemically powerful procedures, lead unnecessarily to social injustice of other kinds such as invidious discrimination against less favored groups by the majority. Retaining the right to vote is one kind of equal status, but if it were to facilitate unjust relations of oppression or hierarchy through unjust political decisions it might be unclear why it has such trumping weight. The question of the substantive justice of the decisions calls for an important role. These approaches have variants, and there are other non-epistemic approaches as well. We put them all aside here for purposes of understanding epistemic approaches, which characteristically make central use of potential epistemic benefits of interpersonal deliberation leading up to political decisions—epistemic deliberation. 9 B. Deliberative and non-deliberative epistemic approaches To understand the importance of interpersonal deliberation to epistemic approaches it is important to first consider epistemic approaches to democracy that make no appeal to it. One simple view, associated with logical critiques stemming from Arrow’s “Impossibility Theorem” (Arrow 1951) is that voting registers and aggregates people’s preferences, resulting in a sound measure of what is in the group’s aggregate interests. Another approach relies on statistical theorems such as Condorcet’s Jury Theorem to argue that a majority is far more likely to be correct than any individual voter, and under minimally favorable conditions highly likely to be correct, none of this depending on any interpersonal discussion or debate (Estlund et al. 1989). Deliberative epistemic approaches, by contrast, would be those that rely at least partly on epistemic benefits of certain forms of public communication and discourse, especially forms involving offering and responding to practical (what to do) and epistemic (what to believe) reasons. If the non-deliberative statistical approach were sound, then an epistemic approach to democracy might not need to query the epistemic value of deliberation. For example, if, even without public deliberation (whatever exactly “no deliberation” would mean) every voter’s chance of choosing the better candidate or law was just a bit better than random, and each voter’s acts are 10 statistically independent of those of other voters, the Jury Theorem’s math alone shows that decisions made by majority rule would be highly likely to be good or correct (where this means only “the better of the two”).4 There is no dispute about the truth of the Jury Theorem as a piece of math, but scholars continue to debate whether the conditions on the applicability of the Jury Theorem are plausibly met in realistic democracies.5 If not, then the mere number of voters might not support any claim of democracy to have epistemic value. Although this important point is often overlooked, it is a departure from that approach to introduce ostensible epistemic benefits of interpersonal deliberation. Below, we concern ourselves only with approaches that appeal to deliberation. C. How epistemic? How epistemically ambitious must such appeals to deliberation be? If the question is why all the people should rule by voting, rather than any subset organized in any other way, there is some pressure to try to show that democratic deliberation outperforms every alternative. On a simple version, certain democratic arrangements could be held to be (actually or potentially) supreme: the most reliably accurate procedure on the relevant questions—better than any single expert or panel of experts, etc. On a more epistemically moderate view, democracy’s epistemic value is held to be, even if not better than all other 11 sources, at least sufficient to justify or require democratic arrangement given certain additional non-epistemic virtues of democracy. Perhaps certain other values would serve to explain why epistemically better arrangements are not required or justifiable. For example, maybe democratic arrangements are necessary to avoid morally undesirable social hierarchy, but would not be permitted if they were not at least epistemically adequate. Or maybe the epistemically best available elite is not widely enough agreed upon to have that distinction, widely enough for stability, or for the liberal moral standard of acceptability to all reasonable points of view. Indeed, whether the alleged epistemic value is supreme or only sufficient, the justification of democracy might be rested on its supposedly true epistemic virtues, or on its epistemic virtues so far as these can be agreed within the politically permissible set of reasons, excluding considerations that are disputable among reasonable citizens. This last idea embeds epistemic deliberative democracy within a form of political liberalism, as developed especially by Rawls. The epistemic potency that is required of deliberation, then, will vary depending on the broader normative theory in which the idea is embedded. D. Epistemic about what? Since the epistemic dimension of a democratic theory is one in which political decisions tend toward being good or correct or true, it is important to say what 12 it is that they might be correct or true about, or in what way they might be good. Some epistemic deliberative democratic theories emphasize the question of arriving at good or correct answers to practical or moral political questions, and this is arguably the main line of epistemic approaches from Rousseau through Rawls, Cohen, Estlund, etc. On that approach, voters might be seen as faced with the question: “Which of the available alternatives best respects or promotes justice?” or “Which of these is the right thing (collectively) to do?” A common complaint about morally construed epistemic approaches of this kind is that, so it is claimed, there is no correct answer to what ought morally to be done in political decisions.6 If this is a sweeping skepticism about moral ideas, or even just moral ideas in political contexts—that they are all mistakes—then it is not any distinctive challenge for epistemic approaches to democracy, since it would deny that there is any answer to the question whether any society at any time ought to be democratic or to respect certain legal rights or engage in wars of conquest, and so on. There are critiques of “moralism” in political philosophy, in favor of an approach sometimes dubbed “realism,” though the terms of the debate are themselves under dispute.7 Suffice it to say that deliberative epistemic approaches to democracy have been content to assume that some political decisions are morally wrong, and some more than others. The aim of such a theory, then, is not to also go deeper in order to refute the outright moral skeptic. 13 A distinct approach to the idea of good or correct outcomes is less focused on correct answers to moral questions, and argues in favor of broadly democratic social (not merely political) arrangements on the ground that they are epistemically favorable for the discovery of good or sound actions and beliefs of all kinds, ranging from morality and politics to science and the arts. Classic antecedents for such views would include John Stuart Mill’s theory of the value of broad social freedom of action and expression, and Dewey’s account of human knowledge as a process of identifying and overcoming problems, a value by no means limited to moral or even political problems. (Anderson 2006, Talisse 2007) E. Some critiques and defenses of epistemic deliberative normative democratic theories There are several common lines of critique of deliberative approaches to democracy, arguments that have nothing in particular to do with the subset that contain an epistemic dimension. They would need to be answered, but they are not especially pertinent to our narrower topic here. Briefly, one obvious line of critique of epistemic approaches is to deny that democratic arrangements really do, or could, have the degree of epistemic value that certain theories would need them to have to render democracy legitimate, authoritative or justified. These debates depend on questions, to be surveyed in Section III, about what the epistemic mechanisms are and how much might be expected of them. But 14 in the background it is important to see that a normative theory is not refuted by the claim that its standards are not likely to be met, unless it (optionally) claims otherwise. As for objections that are more specific to epistemic versions of deliberative democratic theory, we have briefly encountered one of these already, namely that there is no such thing as a procedure-independent standard on the relevant questions, such as what justice requires. A second line of objection is that the epistemic value of democratic deliberation could never be assessed without having independent access to the right answers, in which case the democratic procedure would be epistemically unnecessary. 8 A third line of objection is that focusing on the epistemic as opposed to the intrinsic value of democratic procedures would somehow endanger and perhaps even “disfigure” democracy itself (Urbinati 2015, 81-127). III. Epistemic mechanisms What reasons do we have to believe that democratic deliberation understood as a way of arriving at collective decisions has epistemic properties, i.e., the ability to track a procedure-independent standard of correctness (whatever one may understand by this)? The following aims to give an account of deliberation, and more specifically democratic deliberation, as the epistemic engine of 15 a properly conceived (“deliberative”) democracy. It also aims to clarify the relation of democratic deliberation to consensus and majority rule. A. Epistemic properties of deliberation: As noted at the start, deliberation means, roughly, the pondering and weighing of reasons or an exchange of arguments for or against a given view. In that sense deliberation can refer to an internal dialogue in the vein of “deliberation within” (Goodin 2005), an intersubjective exercise among individuals, or a deliberation occurring among entities larger than individuals, as in systemthinking (Mansbridge and Parkinson 2012). The idea that intersubjective deliberation (leaving aside anything about “democratic” for the moment) has epistemic properties is an old one. It can be traced back all the way to Aristotle’s idea that a “feast to which the many contribute” is better than a feast organized by one person only, all the way to Mill’s emphasis on diversity of points of view in helping the truth overcome falsities and triumph in a free market of ideas. An underlying assumption of these views is that there is a self-revealing nature of the truth, which when made apparent by the exchange of viewpoints is supposed to convince all participants in the deliberation (if not instantaneously, over time, and if not inexorably then at least under favorable conditions). This is something best ex- 16 pressed, perhaps, by Habermas’ idea of the “unforced force of the better argument” (Habermas 1991). How does this “unforced force of the better argument” work in practice? Let us look at the way deliberation functions in a nicely idealized (but not too idealized) model: The deliberations of jurors in the film 12 Angry Men.9 One of the turning points in the deliberation comes when Juror 8 produces a copy of the murder weapon, a cheap switchblade that he said he was able to buy for a fistful of dollars around the corner of the tribunal, disproving at once the unusualness and identifying nature of the weapon. Another argument is produced by Juror 5, who grew up in a violent slum and can explain the proper way of using a switch-blade, raising doubts in the process about the plausibility of the eyewitness’s description. The eyewitness’s reliability is further put in doubt when it becomes clear that she usually wears glasses (as evidenced by red marks on the side of her nose observed by the jurors when she came to testify to the bar). Ultimately a unanimous consensus emerges that the young man should be found not guilty. The story illustrates the epistemic properties of good deliberation. First, it allows participants to weed out the good arguments, interpretations, and information from the bad ones (e.g., the switchblade is not as unique a weapon as previously thought and can only be used a certain way). Second, deliberative problem solving can also produce synergies, that is create new solutions out of 17 the arguments, information, and solutions brought to the table (e.g., making sense of the red marks on the eye-witness’ nose in a way that proves decisive to the interpretation of her reliability). Third, hearing the perspectives of others may entirely reshape a person’s view of the problem and introduce possibilities not initially considered (e.g., the eyewitness testimony cannot be trusted after all). Finally, in the ideal, good deliberation produces unanimous consensus on the “right” solution (“not guilty” in this case). The example also illustrates the specific merit of deliberation among a diverse group of people. In the story all twelve jurors mattered, in all their differences, because it is only through the interplay between their conflicting interpretation of the evidence and arguments—colored as those are by their personal history, socio-economic background, type of intelligence etc.—that something like the truth ultimately emerges. The epistemic properties of deliberation importantly manifest in spite of the fact that the protagonists are far from ideal human beings. One juror just wants to be done with the deliberation and go to a baseball game, one is a bigoted racist, another is biased by irrelevant fatherly emotions… Deliberation, in other words, can overcome a number of moral and cognitive limitations. The logic of epistemic deliberation is well captured by a theorem by Lu Hong and Scott Page, the Diversity Trumps Ability Theorem, which states that under certain conditions10 “a randomly selected collection of problem solvers 18 outperforms a collection of the best individual problem solvers” (Hong and Page 2004: 16388; Page 2007: 163). In other words, “diversity trumps ability” and our twelve angry men are better than twelve clones of, say, juror number 8 (arguably the smartest of the lot) would have been. Diversity here refers to cognitive diversity, which is roughly the difference in the ways different people will think about a problem in the world. On this model, cognitive diversity is not diversity of fundamental values or goals, which would actually harm the collective effort to solve a problem, though it is compatible with degrees of less fundamental value-diversity. This counterintuitive result can be made more comprehensible through the spatial metaphor of the passing of the baton between variously resourceful climbers on a rugged landscape. Whereas smart but homogenously thinking problem-solvers will tend to get stuck at high but local optima, the diverse group is more likely to have members guide each other from lower optima to the global one, because as a group of diverse individuals they explore more of the rugged landscape. It is worth emphasizing that this account of the epistemic logic at work in problem-solving among cognitively diverse groups is distinct from the statistical logic behind the Condorcet Jury Theorem or the Miracle of Aggregation (or another one of Hong and Page’s results, the Diversity Theorem) in that it has nothing to do with the law of large numbers. The point here is not that a clear 19 signal will emerge out of the noise of random errors that cancel out, even though the good and bad input alike get aggregated. It is that deliberation will weed out the bad information and arguments from the outcome entirely. B. Epistemic properties of democratic deliberation While the arguments above may account for the epistemic properties of deliberation among cognitively diverse people, it does not quite justify democratic deliberation in the sense of deliberation that 1) involves all and 2) involves all on an equal standing. Democratic deliberation can indeed be specified as intersubjective deliberation that takes place specifically in a “public” manner “among free and equal individuals” (adapted from Cohen 1989) and is also inclusive of the entirety of the relevant group, though this condition is generally left implicit in a lot of the literature in deliberative democracy.11 Democratic deliberation, in order to count as plausibly democratic, thus requires publicity of the exchanges, full inclusiveness, and equal standing and equal opportunities for participation among participants (‘free and equal’). Theorists appreciative of the epistemic value of deliberation may not necessarily see the epistemic value of democratic deliberation thus understood. Mill, after all, though a deliberative democrat on most readings, was also an advocate of a plural voting scheme that gave more voice (in the form of votes) to the 20 learned. Clearly one can believe in the value of deliberation and not think that all involved should have an absolutely equal right to be heard. Landemore (2012 and 2013) has arguably offered the missing link between the epistemic properties of deliberation and democracy per se, at least when it comes to the inclusive and egalitarian features of democratic deliberation (the publicity element has yet to be shown to have epistemic properties of its own). Landemore argues that more inclusive assemblies are simply more likely to be cognitively diverse. To the extent that cognitive diversity is a key ingredient of collective intelligence, and specifically one that matters more than average individual ability, the more inclusive the deliberation process is, the smarter the solutions resulting from it should be, overall. Numbers, in other words, function as a proxy for diversity (“Numbers Trump Ability Theorem”). Where allinclusiveness is not feasible, a second-best solution is delegation to a randomly selected sample of the group. A key assumption of this argument is the radical uncertainty faced by political decision-makers when it comes to issues of the common good. This fundamental uncertainty (which is an assumption about the world, not necessarily the subjective epistemic stage of the deliberators) is what renders all-inclusiveness on an equal basis epistemically attractive as a model for collective decision-making. Given the complexity of the world, which generates this uncertainty, egalitarian inclusiveness is “ecologically rational” (Landemore 2014). 21 C. Measuring the epistemic quality of democratic deliberation? Empirical research on deliberation was not until recently framed in epistemic terms and to that extent has yet to fully prove or falsify the claims of epistemic deliberative democrats. It was, however, always understood as empirical research about a specifically democratic type of deliberation (as defined above) and thus in what follows we use deliberation and deliberative as short-hands for democratic deliberation and democratically deliberative. There are various ways one could go about measuring the epistemic quality of outputs generated by (democratic) deliberation, from the more indirect to the more direct. A first proxy for the substantive quality of deliberative outcomes is, for example, the objective level of information people have postdeliberation, as compared to their pre-deliberative beliefs and preferences. This is in some respect what Jim Fishkin’s deliberative polls measure (e.g., Fishkin 2009). The presumption here would be that as people’s views are more informed, they are also more likely to be right--although there is of course no guarantee. Another route is to measure the procedural properties of deliberation, as in the Discourse Quality Index (DQI) (Steenbergen et al. 2003; Bächtiger et al. 2010), which codes, among other things, for how equal, respectful, and argumentatively sophisticated people’s speech acts are. Yet other routes are the 22 index of intersubjective consistency (Niemeyer and Dryzek 2007), the index of “cognitive complexity” (Wyss, Beste, and Bächtiger 2015), or the measurement of adequate support for given conclusions (Schaffer and Friberg-Fernros forthcoming). Finally, one may also want to measure the quality of deliberative outcomes in terms of the corresponding’s decisions impact on the world. Are the solutions put forward “validated” by the outside world, i.e., actual empirical success? At a first level of intuition, the fact that democracies—more “deliberative” regimes than their known alternatives-- have been doing well by multiple standards both in Ancient Greece (Ober 2008) and over the last 250 years— even causing economic growth on some readings of the available evidence (Acemoglu et al. 2014)--would seem to at least not contradict this prediction. Small-scale lab experiments involving problems with mathematical, logical, factual, or otherwise uncontroversial answers also support the case that deliberating groups solve riddles faster and better than less deliberative ones (Clément, Krause, von Engelhardt, Krause, and Kurvers 2013). Experimental evidence obtained in the developing world, finally, suggests that deliberation promotes a certain number of uncontroversially good outcomes, such as efficiency (Goeree and Yariv 2011) or the reduction of clientelism (Fujiwara and Wantchekon 2013). 23 Should (democratic) deliberation aim at consensus/unanimity? The ideal of consensus as unanimous agreement on a decision resulting from a deliberative process lies at the heart of early theories of deliberative democracy (Cohen 1986: 22; Habermas 1991: 66; Elster 1986: 112; Young 1996: 122) Does consensus form a promising normative horizon for epistemic deliberation? In order to answer this question, Landemore and Page (2014) suggest to first distinguish between at least three meanings of consensus as a normative ideal: consensus as (i) a goal, (ii) a stopping rule, or (iii) an outcome. While consensus can also be two or three of these things at once, it need not be and these analytical distinctions clarify the debate. First, consensus can be a goal— that is, a direct aim that deliberators seek or should seek to achieve when discussing with each other--in contrast with, say, truth or the promotion of certain interests. Second, consensus can also be interpreted as a stopping rule, that is, the rule by which deliberation is brought to an end and a group decision considered taken. On that interpretation, consensus is equivalent to unanimity rule and to be contrasted with other stopping-rules, such as simple majority rule or super majority rule. Lastly, consensus can be interpreted as an ideal outcome—that is, the result of an ideal deliberation. Being an outcome rather than a goal, consensus in that sense is not something that deliberators necessarily pursue directly. It happens instead as a by-product of something else, like pursuing the truth. 24 Distinct lines of criticism questioning consensus can be identified in light of these distinctions. The first criticism of consensus is aimed at the stopping rule that requires unanimity—equivalent to giving each individual a veto and thus giving undue weight to minorities’ preferences (see Rae 1975 and McGann 2006). Another line of criticism objects to consensus as a direct goal of deliberation, because rational consensus may be hard in practice to distinguish from compromise (Steiner et al. 2004); because defining consensus as a goal distorts incentives for participants in a deliberation, creating pressure to reach an agreement (e.g., Mackie 2006, 285); or, conversely, because defining consensus as a goal invites strategic conformism12 (Feddersen and Pesendorfer 1998). These criticisms suggest that the only way consensus should ever be considered a “goal” of deliberation is in the sense of being an indirect goal of deliberators, namely a hoped-for by-product of their argumentative exchanges, while they are directly aiming for something else, such as the truth or a better understanding of the issues (Fuerstein 2014). The third, most important line of criticism, however, is directed at consensus as the ideal outcome of deliberation. Gutmann and Thompson (2005) thus argue that there is no agreement among deliberative democrats that consensus should be the goal of deliberation (by which they mean the “indirect goal” of deliberators or their ideal “outcome”). Indeed, some claim that at the level of principles, other components of the deliberative democracy ideal, such as rea- 25 son and public justification, point away from rational consensus, not towards it (e.g., Gaus 1997, 207). Still other objectors point out that when it is reached, consensus as an outcome often signals something less than ideal, for example polarization—a post-deliberative reinforcement of previously held beliefs— rather than rational consensus (see e.g. Mendelberg and Oleske 2000; Sunstein 2003). While not all of the objections to consensus prove decisive, their cumulative effect has been to generate discomfort with the ideal of consensus among deliberative democrats (a discomfort shared early on by some philosophers, e.g., Rescher 1993). Symptomatically, influential deliberative democrats have recently gone back to embracing the full legitimacy of stopping-rules for deliberation that used to be considered regrettable second-best of consensus, such as majority rule or even the kind of non-communicative agreement reached through bargaining (e.g., Mansbridge et al. 2010). In this approach, the ideal termination of deliberation is not agreement but disagreement, followed by a non-deliberative decision rule. It is possible, however, that the epistemic appeal of consensus will vary depending on the context and task at stake. In problem solving contexts, consensus as an ideal outcome of deliberation retains an epistemic appeal as a “marker” of truth, signaling that no one knows or can construct a better idea. In predictive contexts, however, consensus has almost no normative appeal as a 26 stopping rule and little normative appeal as an outcome. Instead, when relatively equally compelling logic and evidence support multiple models, the group members should cultivate “positive dissensus,” a form of disagreement that is epistemically beneficial for the group and ultimately leads to more accurate aggregated predictions (Landemore and Page 2014). D. Deliberation and majority rule Deliberation and majority rule have sometimes been pitched as rival mechanisms for decision-making, one being championed by the deliberative democracy camp, the other by so-called aggregative democrats. Recent research in epistemic democracy tends to portray them as complementary decisionprocedures and emphasize their distinct epistemic properties. Whereas deliberation is more adapted to pure problem-solving contexts, majority rule is a faster and more accurate tool for purely predictive tasks. E. Empirical objections: Some critics have suggested that intersubjective deliberation (democratic or not) may in fact either have no tendency to improve outcomes, or even a tendency to make them worse. Robert Goodin and Simon Niemeyer (2003) have thus suggested that “deliberation within” is what is doing all the work in observed deliberations, as opposed to the intersubjective deliberation going on in 27 various mini-publics that they observed. For them, it is the internal pondering of reasons that people engage in when reading briefing material that changes people’s minds, not the exchange arguments that takes place in small groups or plenary discussions. However, given the likely evolutionary reasons behind human use of reasoning and the well-documented motivation bias of individual reasoning, it can be argued that deliberation is more likely to produce the truth when it is intersubjective, i.e., social, rather than internal (Mercier and Landemore 2012). Cass Sunstein has made the stronger argument that deliberation can make things worse because the “law of polarization” dooms even slightly likeminded groups to become more entrenched in their pre-deliberative beliefs. Deliberation as celebrated by deliberative democrats, Sunstein concludes, is overrated. In his view, the underlying mechanisms of group deliberation “do not provide much reason for confidence” (Sunstein 2002, 187). He further suggests that in the context of groups engaged in so-called “enclave deliberation” (Mansbridge 1994), one is better off aggregating judgments as they are rather than make things worse by encouraging people to talk things out. The gist of the answers to Sunstein so far is to deny that the “talk” that went on in the experiments he reports on amount to genuine deliberation, because participants fail to engage conflicting arguments as opposed to merely diverse one (e.g., Manin 2005, 9; see also Thompson 2008, 502 and Lan- 28 demore 2013, 123-142 for similar points). It has also been argued that while Sunstein’s evaluation of group discussion is too pessimistic, his evaluation of prediction markets and internet devices is too optimistic, markets being imperfect and the internet being vulnerable to astroturfing by the powerful and wealthy (Mackie 2009). Others conjecture that the difference between talk that polarizes and talk that does not lies in the enforcement by trained moderators of deliberative norms such as speaking one’s mind, listening to others, behaving respectfully, and learning and persuading others through reasons. This conjecture is supported by experimental results (Grönlund, Herne & Setälä 2015). Discussion following deliberative norms arguably reverses polarization tendencies within like-minded groups (Grönlund et al. forthcoming). Conclusion As we have seen, the topic of epistemic democratic deliberation has gained its prominence in recent democratic theory partly from close interpretations of the seminal normative works in the broader research paradigm of “deliberative democracy,” such as Cohen and Habermas, and partly out of critiques of normative democratic theories emphasizing certain purely procedural values such as procedural fairness or other symmetrical ruling relations. But the underlying issue of how well democracy can perform is, and always has been, also central 29 to a reflective political engagement in broadly democratic culture. As we write in 2016, though this is nothing new, just as some ostensible political disasters are chalked up to insufficiently democratic political procedures (for example, critiques of widening material inequality), others are put down to an excess of democracy (such as the rise of Donald Trump or other demagogues, or abrupt momentous changes by referendum such as "Brexit."). Are some forms of politics to be preferred to others on grounds of their being expected to make better decisions at all, or is that issue (somehow) to be put aside? If it is not put aside, is democracy the site of a separate value that might compete with such an epistemic criterion? Is democracy to be recommended partly on epistemic grounds, and if so, what institutional features of democratic arrangements might drive the epistemic value, and by what mechanism? And is the epistemic democracy paradigm wholly aspirational, at best, or are real political choices already properly informed by these matters? If no version of democracy could perform as well as some specified non-democratic alternative, how are the non-epistemic procedural values that democracy might instantiate (or is that just as unrealistic?) to be weighed against the values by which the alternative would perform better—values ranging from matters of basic health, welfare, education, distributive and relational equality, antibigotry, equitable infrastructure, cultural climate, limiting state and non-state violence, incarceration, economic monopoly, accumulation of economic and 30 political power, and much more? Procedural fairness and equality plausibly have some intrinsic value, but so do outcomes that are up for political decision (or political default, as in an anti-state position), and they are presumably not easily outweighed. If democracy is held to have an epistemic response to this challenge, it needs to rely at least partly on epistemic effects of interpersonal, public, democratic deliberation, a set of topics on which there is much work being done, and much left to do. References Acemoglu, D., Naidu,S., Restrepo, P. and Robinson, J.A. 2014. Democracy Does Cause Growth. NBER Working Paper No. 20004. doi: http://www.nber.org/papers/w20004 Anderson, E. 2006. The Epistemology of Democracy. Episteme: A Journal of Social Epistemology, 3: 8–22. Arrow, Kenneth, J. 1951. Social Choice and Individual Values. New Haven: Yale University Press. 2nd Edition published 1963. Bächtiger, A., Pedrini, S. and Ryser, M. (2010). Prozessanalyse politischer Entscheidungen: Diskurstypen und Sequenzialisierung. Jahrbuch für Handlungs- und Entscheidungstheorie. Band 6: Schwerpunkt Neuere Entwicklungen des Konzepts der Rationalität und ihre Anwendungen, Wiesbaden: VS Verlag für Sozialwissenschaften: 193-226. Brennan, J. 2016 Against Democracy. Princeton: Princeton University Press. Clément, R.J.G., Krause, S., von Engelhardt, N., Faria, J.J., Krause, J. and Kurvers R.H.J.M. 2013. Collective Cognition in Humans: Groups Outperform Their Best Members in a Sentence Reconstruction Task. PLoS ONE 8(10): e77943. Christiano, T. 1996. The Rule of the Many: Fundamental Issues in Democratic Theory. Boulder, C.O.: Westview Press. Cohen, J. 1986. An Epistemic Conception of Democracy. Ethics, 97: 26-38. 31 Coleman, J. and Ferejohn, John. 1986. Democracy and Social Choice. Ethics, 97: 6–25. Elster, J. 1986. The Market and the Forum. In Foundations of Social Choice Theory. eds. Elster, J. and Hylland, A. Cambridge: Cambridge University Press. Estlund, D. 1994. Opinion Leaders, Independence, and Condorcet's Jury Theorem. Theory and Decision, 36: 131-162. Estlund, D. 1997. Beyond Fairness and Deliberation: The Epistemic Dimension of Democratic Authority. Pp. 173-204 in Deliberative Democracy: Essays on Reason and Politics, eds. Bohman, J. and Rehg, W. Cambridge, MA: MIT Press. Estlund, D. 2008. Democratic Authority: A Philosophical Framework. Princeton: Princeton University Press. Estlund, D., ed. 2008. Episteme 5 (1). Estlund, D.M., Waldron, J., Grofman, b. and Feld, S. 1989. Democratic Theory and the Public Interest: Condorcet and Rousseau Revisited. The American Political Science Review, 83: 1317–1340. Farrar, C., Fishkin, J., Green, D., List, C., Luskin, R. and Paluck, E. 2010. Disaggregating Deliberation’s Effects: An Experiment within a Deliberative Poll. British Journal of Political Science, 40: 333–47. Fishkin, J. 2009. When the People Speak: Deliberative Democracy and Public Consultation. Oxford: Oxford University Press. Fishkin, J. S., & Luskin, R. C. 2005. Experimenting with a democratic ideal: Deliberative polling and public opinion. Acta Politica, 40: 284–298. Feddersen and Pesendorfer, 1998. “Convicting the Innocent: The Inferiority of Unanimous Jury Verdictsunder Strategic Voting.” American Political Science Review 92, no. 1: 23–35. Fujiwara and Wantchekon. 2013. “Can Informed Public Deliberation Overcome Clientelism? Experimental Evidence from Benin.” American Economic Journal: Applied Economics, 5(4): 241–255. Fuerstein, Michael. 2014. Democratic Consensus as an Essential Byproduct. The Journal of Political Philosophy, 22(3): 282-301. Gastil, J. and Dillard, J. P. 1999. Increasing political sophistication through public deliberation. Political Communication, 16: 3-23. Gaus, Gerald. 1997. Reason, Justification, and Consensus: Why Democracy Can’t Have It All in Bohman, J. and Regh, W. Deliberative Democracy: Essays on Reason and Politics. Cambridge, MA: M.I.T. Press. Goeree, J.K. and Yariv, L. 2011. An Experimental Study of Collective Delibera- 32 tion.” Econometrica, 79: 893–921. Goodin, R.E. 2000. Democratic Deliberation Within. Philosophy & Public Affairs, 29: 81-109. Goodin, R.E. and Niemeyer, S.J. 2003. When Does Deliberation Begin? Internal Reflection versus Public Discussion in Deliberative Democracy.” Political Studies, 51: 627–49. Grofman, B. and Feld, S.L. 1988. Rousseau's General Will: A Condorcetian Perspective. The American Political Science Review, 82: 567–576. Grönlund, K., Herne, K., and Setälä, M. 2015. Does Enclave Deliberation Polarize Opinions? Political Behavior, 37(4): 995-1020. Grönlund, K., Himmelroos, S., and Strandberg, K. Unpublished. “Can Deliberative Norms Prevent Group Polarization in Like-Minded Groups?” Guttmann, A. and Thompson, D. 1996. Democracy and Disagreement: Why Moral Conflict Cannot Be Avoided in Politics and What Should Be Done About It. Cambridge, MA: Harvard University Press. Gutmann, A. and Thompson, D. 2004. Why Deliberative Democracy? Princeton, NJ: Princeton University Press. Habermas, J. 1991. Moral Consciousness and Communicative Action. Cambridge, MA: MIT Press. Hong, Lu and Scott E. Page. “Groups of diverse problem solvers can outperform groups of high-ability problem solvers.” Proceedings of the National Academy of Science vol. 101 no. 46: 16385–16389 Landemore, H. 2012. “Democratic Reason: The Mechanisms of Collective Intelligence in Politics.” In Landemore and Elster, Collective Wisdom: Principles and Mechanisms. Cambridge: Cambridge University Press. Landemore, H. 2013. Democratic Reason: Politics, Collective Intelligence, and the Rule of the Many. Princeton: Princeton University Press. Landemore, H. and Page, S.E. 2014. “Deliberation and Disagreement: Problem Solving, Prediction, and Positive Dissensus.” Politics, Philosophy, and Economics, 14(3): 229-254. Luskin, R. C. 2002. “From Denial to Extenuation (and Finally Beyond): Political Sophistication and Citizen Performance.” Pp. 281–305 in Thinking about Political Psychology, ed. Kuklinski, J.H. New York: Cambridge University Press. Mackie, G. 2006. Does democratic deliberation change minds? Politics, Philosophy, and Economics, 5(3): 279-303. Mackie, G. 2009. Astroturfing Infotopia. Theoria, 56: 30-56 33 Manin, B. 2005. Deliberation: Why we should focus on debate rather than discussion.” Paper presented at the Program in Ethics and Public Affairs Seminar. Available at: http://www.princeton.edu/csdp/events/Manin101305/Manin101305.pdf Mansbridge, J. 1994. Using Power/Fighting Power. Constellations 1(1): 53–73. Mansbridge, J., with James Bohman, Simone Chambers, Andreas Föllesdal, Archon Fung, Cristina Lafont, José Luis Martì. 2010. The Place of SelfInterest and the Role of Power in Deliberative Democracy. Journal of Political Philosophy 18(1): 64-100. Mendelberg, T. and Oleske, J. 2000. Race and Public Deliberation. Political Communication, 17: 169-191. Mercier, H. and Landemore, H. 2012. Reasoning is for arguing: Understanding the successes and failures of deliberation. Political Psychology 33(2): 243-258. Niemeyer, S. and Dryzek, J. 2007. The Ends of Deliberation: Meta-consensus and Inter-subjective Rationality as Ideal Outcomes.”Swiss Political Science Review 13(4): 497-526. Ober, J. 2008. Democracy and Knowledge: Innovation and Learning in Classical Athens. Princeton: Princeton University Press. Page, Scott E. 2007. The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies. Princeton: Princeton University Press. Pettit, P. 2012. On the People’s Terms. Cambridge: Cambridge University Press. Rae, D. 1975. The Limits of Consensual Decision. American Political Science Review 69(4): 1270-1294. Rawls, J. 1971. A Theory of Justice. Cambridge, MA: Harvard University Press. Rawls, J. 1985. Justice as Fairness: Political not Metaphysical. Philosophy & Public Affairs 14 (3): 223-251. Raz, J. 1990. Facing Diversity: The Case for Epistemic Abstinence. Philosophy and Public Affairs 19(1): 3-46. Rescher, N. 1993. Pluralism: Against the Demand for Consensus. New York: Oxford University Press. Riker, W. 1982. Liberalism Against Populism. London: W.H. Freeman & Co Ltd. Schaffer, J. K. and Friberg-Fernros, H. Forthcoming. Assessing the epistemic quality of democratic outcomes in terms of adequate support for conclusions. Journal of Social Epistemology. 34 Rossi, E. and Sleat, M. 2014. Realism in Normative Political Theory. Philosophy Compass 9: 689–701. Schwartzberg, M. 2015. Epistemic Democracy and its Challenges. Annual Review of Political Science 18: 187–203. Sunstein, C. 2002. “The Law of Group Polarization.” Journal of Political Philosophy 10 (2): 175–95. Steiner, J. and Dorff, R.H. 1980. Decision by Interpretation: A New Concept for an Often Overlooked Decision Mode. British Journal of Political Science, 11 (1): 1– 13. Steenbergen, M. R., Bächtiger, A., Spörndli, M. and Steiner, J. 2003. Measuring Political Deliberation: A Discourse Quality Index. Comparative European Politics, 1: 21-48. Talisse, R.B. 2007. A Pragmatist Philosophy of Democracy. New York: Routledge. Talisse, Robert. 2012. “Deliberation.” In The Oxford Handbook of Political Philosophy, ed. David Estlund, Oxford University Press. Thompson, D. 2008. Deliberative democratic theory and empirical political science. Annual Review of Political Science, 11: 497–520. Urfalino, P. 2010. Deciding as Bringing Deliberation to a Close. Social Science Information 49(1): 109-138. Urbinati, N. 2015. Democracy Disfigured. Cambridge: Harvard University Press. Waldron, J. 1999. Law and Disagreement. Oxford: Oxford University Press. Wyss, D., Beste, S. and Bächtiger, A. Forthcoming. A Decline in the Quality of Debate? The Evolution of Cognitive Complexity in Swiss Parliamentary Debates on Immigration (1968-2014). Swiss Political Science Review. Young, I. M. 1996. “Communication and the Other: Beyond Deliberative Democracy.” In Democracy and Difference . Edited by Seyla Benhabib. Princeton, NJ: Princeton. Young, I. M. 2000. Inclusion and Democracy. Oxford: Oxford University Press. NOTES 35 1 Contrary to a common critique of "deliberative democracy," none of this implies that democratic deliberation must take the form of explicit appeals principles, logic, etc. Young (e.g., 1996) has emphasized the importance of other modes of expression such as stories, songs, and performances, and deliberative accounts of democratic theory can (arguably) agree. For an instructive critical discussion of the evolution of this strand of Young’s thought, see Talisse (2012). 2 Brian Barry was an early interpreter of Rousseau as having, in part, an epis- temic conception of democracy, including nascent reference to Jury Theorem reasoning. See Barry Political Argument, (Routledge, 1965) Grofman and Feld (1988) developed the Jury Theorem reading of Rousseau. See replies by Estlund and Waldron, with a rejoinder by Grofman and Feld (Estlund et al. 1989). 3 For a sampling see special issue of the journal, Episteme devoted to “Epistem- ic Approaches to Democracy,” (Estlund ed. 2008) and references therein. 4 For an accessible sketch of the proof, see Estlund 1994, 132-37. 5 For a critique of the Jury Theorem’s use in democratic theory, see Estlund 2008, chapter 12, “The Irrelevance of the Jury Theorem.” For a more detailed treatment of the theorem and its prospects in democratic theory see Goodin, Reflective Democracy, (Oxford 2003), esp. Ch. 5, and his discussion of the Bayesi- 36 an variant in Ch. 6. Both Estlund and Goodin contain reference to further sources as well. 6 “…surely democracy must be largely occupied with questions that are not plausibly truth-apt” (Schwartzberg 2015, 199). 7 For a review of recent literature see Rossi and Sleat 2014. 8 Waldron 1999, 252-54. 9 Here we borrow from Landemore, op. cit. 10 There are four distinct conditions for the Diversity Trumps Ability Theorem to apply (Page 2007, 163). Whether or not they all translate neatly to the real world of politics and democratic citizens is a contested issue. See Anderson 2006 and Landemore 2013 for application to the democratic context. 11 This is less true of Habermas (1996) and Young (2000) who both make use of the all affected interests principle to define the appropriate scope of inclusion. Neither, however, theorizes this condition with the degree of elaboration it probably deserves. 12 Aiming for unanimity gives individuals an incentive to vote against their preferences to produce what, given their information, is likely to be the better outcome. 37