The instrumental value of deliberative democracy – or, do we have good reasons to be deliberative democrats?

Though commanding a prominent role in political theory, deliberative democracy has also become a mainstay of myriad other research traditions in recent years. This diffusion has been propelled along by the notion that deliberation, properly conceived and enacted, generates many beneficial outcomes. This article has three goals geared toward understanding whether these instrumental benefits provide us with good reasons – beyond intrinsic ones – to be deliberative democrats. First, the proclaimed instrumental benefits are systematized in terms of micro, meso, and macro outcomes. Second, relevant literatures are canvassed to critically assess what we know – and what we do not know – about deliberation’s effects. Finally, the instrumental benefits of deliberation are recast in light of the ongoing systemic turn in deliberative theory. This article adds to our theoretical understanding of deliberation’s promises and pitfalls, and helps practitioners identify gaps in our knowledge concerning how deliberation works and what its wider societal implications might be.

Although deliberative democracy is now the most vibrant area of scholarship in political theory (Pateman, 2012), its reach extends far beyond. Today the theory has a prominent role in myriad other research traditions including (but not limited to) international relations, comparative politics, public administration, law, psychology, ethics, clinical medicine, planning, policy analysis, ecological economics, sociology (especially social movement studies), environmental governance, and communication studies.
At its core, deliberative democratic theory is a normative enterprise (Habermas, 1996;Thompson, 2008). Through equal and non-coercive deliberation between affected individuals, law and public policy gains its legitimacy. Deliberative democracy embodies and discharges a fundamental duty of equal respect for the opinions and interests of others that ground democratic decision-making in the context of pervasive disagreement. Mutual reason giving, then, should be pursued as it forms the basis for justifying the political and social orders that structure our collective lives (Gutmann and Thompson, 1996). On this widespread view, deliberative democracy has intrinsic normative worth: the process of deliberation is valuable for its own sake. This intrinsic normative core has been largely responsible for the enduring significance of deliberative democracy in political theory.
While many proponents still seek to justify deliberative democracy on deontological grounds, these kinds of arguments are now routinely supplemented or supplanted by teleological claims. In this vein theorists and practitioners couch their justifications for deliberative democracy in consequentialist terms whereby the value of the theory lies in its ability to produce certain normatively desirable outcomes. It is these instrumental effects that have been crucial to the diffusion of deliberative democratic theory to other disciplines.
To the extent that deliberative democracy is expected to produce desirable outcomes, these instrumental effects are susceptible to empirical testing. Although evaluating deliberative theory is complicated by a range of factors (Mutz, 2008), much recent scholarship has focused precisely on undertaking this task. Qualitative, quantitative, and experimental methods are being employed to test how, if at all, deliberative democracy produces valuable outputs. Do these empirical studies furnish us with good reasons to be deliberative democrats?
In this article, I address this question by systematizing the existing literature on the instrumental effects of deliberation across three dimensions: micro-, meso-, and macro-level changes. I then document what we knowand what we do not knowabout these instrumental effects from a range of disciplines. 1 The key finding is that we now have good evidence that deliberation produces some desirable consequences. In this sense, deliberative democracy has moved beyond being a 'working theory' to a normative theory with real empirical bite (Chambers, 2003). However, two issues remain critically underexplored. First, limited work has sought to break deliberation into its constitutive features and explore which element produces which effect. As the 'unitary model' of deliberation is questioned, this will be an important line of enquiry (Bächtiger & Beste, 2017;Goodin, 2005). Second, the scope conditions that promote or inhibit good outcomes are also underdetermined. Again, as we begin to probe how actors and fora can be combined across institutional landscapes to produce normative goods, understanding the scope conditions of deliberative benefits is imperative. I expound these limitations, highlight some methodological concerns in addressing them, and establish new directions for research. Third, I analyse these instrumental effects in light of the systemic turn that has dominated deliberative theory in recent years. I discuss how the micro, meso, and macro levels can fit together to provide a robust deliberative system, and reflect on some core tenets of this turn. The final section concludes.
In sum, the article provides a systematic review of the instrumental effects of deliberation relevant to both theorists and practitioners. For theorists, the article sheds light on how deliberation might be understood at different levels, unpacks how the disaggregation of deliberation connects to other normative goals, and helps rethink the role of deliberation in wider systems. For practitioners, the article identifies knowledge gaps concerning how deliberation works at the individual, group, and societal levels, directs attention toward the scope conditions of deliberation, and thus suggests directions for empirical research.

The Ideal of Deliberative Democracy and the (Supposed) Instrumental Effects
There are many ways to understand the ideal of deliberative democracy. And befitting its prominence, supporters and critics pick up on different elements when utilizing the concept. Despite this diversity, it is possible to extract several key elements. Centrally, deliberation is a 'talk-centric' mode of decisionmaking that eschews coercive power relations in favour of reason-giving (Chambers, 2003). Deliberators should ideally: foster inclusive and egalitarian interactions; sincerely link reasons to arguments (display justificatory rationale); orient claims toward the common good; frame arguments in terms of reasons acceptable to others (demonstrate reciprocity); show respect, and; be prepared to change their mind when confronted by 'better arguments'. 2 To the extent that decisions, policies, and laws track these desiderata, democratic legitimacy obtains. While this normative vision is foundational to deliberative theory, early work in the field blended intrinsic claims with instrumental ones (Rosenberg, 2005). As Tali Mendelberg (2002) notes, deliberationproperly conceived and enacted -is not just about the process, but is also supposed "to produce a variety of positive democratic outcomes".
Many, if not all, prominent deliberative theorists ascribe to something like this view. For instance, Joshua Cohen (1989) and Jon Elster (1986) contend that deliberative democracy helps to arrive at a rationally motivated consensus. 3 Jane Mansbridge (1995) claims that participation in deliberation makes better citizens. Seyla Benhabib (1996) and Bernard Manin (1987) suggest that deliberation generates broad popular support even under conditions of disagreement. Amy Gutmann and Dennis Thompson (1996) submit that deliberation produces tolerance for the views of others. Simone Chambers (1996) notes how deliberation enables citizens to better grasp their own preferences, to understand shared problems, and to shape decisions in more informed ways. Iris Marion Young (2002) suggests that deliberation portends many virtues, "such as promoting cooperation, solving collective problems, and furthering justice" (p. 26). And the list goes on.
From these early theoretical statements, scholars have begun documenting and assessing the empirical validity of specific instrumental claims. Mendelberg (2002) reviews previous work and argues that deliberation is supposed to lead to enhanced empathy, enlightened preferences, the resolution of deep conflict, engagement in politics, faith in the basic tenets of democracy, perceived legitimacy of political systems, and a healthier civic life. Her review, covering scholarship in political theory as well as social and cognitive psychology, turns up mixed results for deliberation in terms of heightening cooperation, mitigating in-group bias, undercutting power differentials, and producing common understanding across difference. Yet the main result is that benefits are contingent on circumstance, personality, and context.
Diana Mutz (2008) provides a veritable shopping list of supposed desirable outcomes linked to good deliberative processes. Inter alia, these include: awareness of oppositional arguments; tolerance; perceptions of legitimacy; knowledge/information gain; empathy; willingness to compromise; civic engagement; opinion consistency; faith in democratic institutions; consensual decision-making; social capital and trust, and; depth of understanding of one's own views. Her main argument is that more middle-range theorizing is required to test these benefits systematically. In another review, Shawn Rosenberg (2005) argues pooling the ideas of different participants and subjecting these ideas to collective, reasoned deliberation is thought to yield more knowledgeable citizens, consistent preferences, enhanced emancipation, internal reflection, improved collective decisions, and the bridging of existing social cleavages. Similar to Mutz, his conclusion is that more empirical testing is needed, especially in ways sensitive to existing social and psychological research.
As deliberative democracy reaches into new fields with different theoretical debates and empirical insights, the list of (alleged) instrumental benefits has expanded. Based on previous reviews and an inductive analysis of recent work, I suggest that the instrumental effects of deliberative democracy can be categorised across three dimensions: micro-(individual), meso-(group), and macro-(polity) level changes. Each level can be further disaggregated into three dimensions. For the micro-level, deliberation is said to structure preferences and shift opinions, increase knowledge, and enhance civic participatory desires. At the meso-level, deliberation may foster social learning, undercut polarization, and generate consensus. At the macro-level, deliberation can bolster popular support, help overcome deep divisions in society, and facilitate democratization. This list is not exhaustive, but it does cover deliberation's promises: that is, the ability to make just and democratically-legitimate decisions. 4 Turning now to the three dimensions, I focus on key pieces in different literatures to ensure depth is not substituted for breadth.
The State of the Art: What Do We Know, and What Don't We Know?

Micro-Changes: Preferences and Opinions, Knowledge, and Civic Participation
Most knowledge concerning how deliberation impacts the individual comes from experimental treatments in which individuals are separated into treatment and control groups, measured on a range of variables, engaged in a deliberative process (for those in the treatment), and then re-tested at the end to see if shifts have occurred across one or several of these variables. The main result from different literatures is that deliberation does drive individual-level changes.
4 Several important (alleged) benefits are not included here. Perhaps most notable is deliberation's epistemic, or truth-tracking, potential. Despite this important literature, I leave this aside here due to epistemological concerns. Studying epistemic correctness empirically would necessitate a procedure-independent standard of rightness, as procedural-correctness would fail a standard of falsifiability. Given many situations may have several reasonably correct alternatives, and that sometime the best argument only emerges through deliberation, I simply note this as avenues for future research.
Perhaps the most commonly claimed benefit of deliberation is that the process alters peoples' preferences and opinions in desirable ways, reflecting the emergence of common understanding and agreement on correct courses of action (Chambers, 2003). Much empirical work has focused on substantiating how, or even if, these benefits accrue. Initial work turned up ambivalent findings because opinion changes were measured at the group level which masked individual shifts (Barabas, 2004). However, recent literature has borne out this instrumental benefit by showing that deliberation does structure individual preferences and alter opinions in positive ways (Baccaro, Bächtiger, & Deville, 2016;Goodin and Niemeyer, 2003;Luskin, Fishkin and Jowell, 2002;). 5 This follows the theoretical prediction of John Dryzek and Christian List (2003) who argued that deliberation could shift individual preferences toward singlepeakedness, helping to reduce cycling and therefore undercutting classic social choice dilemmas. This hypothesis has been confirmed in a variety of deliberative poll experiments (Farrar, Fishkin, Green, List, Luskin, & Paluck, 2010;List, Luskin, Fishkin, & McLean, 2013;). While these studies are instructive, it often becomes difficult to pin down what drives these results: what element of deliberation leads to change, and what in the corpus of activity comprising a deliberative poll (recruitment, interviews, moderation, on-site deliberation, post-deliberation surveys, etc.) alters preferences in this way (Farrar et al., 2010). Because deliberative polls tend to measure pre-and postdeliberative preferences, these issues are left in a 'black box' (Bächtiger and Parkinson, forthcoming).
Several scholars have recognizedand sought to correctthis issue. Gerber, Bächtiger, Fiket, Steenbergen, & Steiner, (2014) exploiting the data from the EuroPolis deliberative polluse the discourse quality index (DQI) to find that citizens are able to craft fairly sophisticated deliberative arguments by invoking arguments linked to common good orientation, respect, empathy and inquisitiveness. This does lead to stronger preference change, but the result is tempered by uneven distribution of deliberative capacity across demographic groups (most notably class and geographic origin). 6 Employing evidence from a different minipublic in Finland, Staffan Himmelroos and Henrik Serup Christensen (2014) find that opinion change in deliberators comes down to justificatory rational, respect, and reflection (see also Baccaro et al., 2016). Sean Westwood (2015), in an exemplary study of a deliberative poll concerning the 2004 US primaries, shows that justified statements aimed at persuading individual listeners is the strongest predictor of individual opinion change.
5 For a dissenting view, see Sanders (2012). Using individual-level multivariate panel data from the Europolis Poll he finds that opinions do shift but 'strength of argument' is not causal. However, his reliance on self-perception dilutes this result. 6 See also Gerber, Bächtiger, Shikano, Reber, and Rohr (2016). Siu (2008) also shows that the ability to engage in high quality arguments does not vary across gender. Suiter, Farrell, and O'Malley (2016) show that opinion change is most likely in deliberators under 65 with median knowledge levels.
However, Westwood does not unpack which kind of justifications generate this result. For instance, we should care whether logically consistent arguments, those that attach to common good justifications, or those that display reciprocity generate the most change.
A closely related strand of literature has focused on knowledge gain as a desirable outcome of deliberation (Barabas, 2004). The basic idea here is that the process exposes deliberators to new views, piques their interest, and leads to increased knowledge on this issue. The connections between knowledge gain and preference/opinion change are complex. On many accounts, knowledge gain is treated as an intervening variable that impacts opinions and preferences.
As such not all studies interested in deliberation and knowledge gain treat the latter as a dependent variable (Sanders, 2012). 7 Yet there is a substantial body of work that shows that knowledge gain is a by-product of deliberation. Delli Carpini (1997) demonstrates that deliberation enhances knowledge on that issue, and Cook and Jacobs (1998) found that deliberation in a forum concerning social security led to increased knowledge on the program and plans. Gastil (2008) likewise finds that participation in deliberative forums raises individuals' levels of interest in the discussion topic, including how frequently they seek knowledge on that issue, but does not elucidate a causal mechanism behind this shift. Early scholarship thus turned up the conclusion that discussing an issue made individuals more knowledgeable about that topic, including deepening their own position and gaining knowledge of alternate views (Barabas, 2004;Rosenberg, 2005).
Building upon these finding, scholars have moved toward unpacking the conditions under which knowledge is obtained. Farrar et al. (2010) find that deliberation enhances individual knowledge significantly (in the statistical sense) when issue salience is low: individuals have less established views and therefore more 'room to move'. Andersen and Hansen (2007) employing panel data from a deliberative poll on the Danish referendum to adopt the Euro show that knowledge is enhanced through deliberation on the subject, but that initial deliberation has a stronger effect than later deliberation. This is exactly what scholars of path-dependence would predict, but it remains to be seen whether some form of arguments (justificatory, reciprocal, generalizable, etc.) given later in a process can undo earlier arguments that are less well constructed. Kimmo Grönlund and his collaborators (2010) have also looked at whether deliberative quality increases knowledge. They find that when groups are required to form a common statement at the end of deliberation, these groups both deliberate 'harder' (i.e. weighing alternate positions more) and generate more knowledge gains (Grönlund, Setälä, & Herne, 2010). To date, I know of no longitudinal study that seeks to understand whether deliberation generates longer lasting forms of knowledge gain, or through what specific mechanism.
The final individual-level change brought about by deliberation involves enhanced desires for civic participation or engagement. This outcome was treated sceptically by early research in the field. Canonically Mutz (2002) found that exposure to more plural media increased political engagement, but exposure to a greater number of opposing discussions decreased willingness to engage in politics. 8 Yet other work has determined a connection between deliberation and an individual's civic participatory desires. For example, Fishkin and Luskin (2005) found that deliberative practices enhance the skills associated with citizenship. Andersen and Hansen (2007) and Christensen, Himmelroos, and Grönlund (2017) show that deliberation increases willingness to participate in deliberative bodies, but may discourage participation in formal representative politics. Neblo, Esterling, Kennedy, Lazer, and Sokhey (2010), in an impressive lab-inthe-field experiment, come to an alternate conclusion: those less interested in traditional politics are most likely to be motivated to engage in political participation after deliberation. Finally Knobloch and Gastil (2015), employing longitudinal survey data from two deliberative forums, find that deliberation makes individuals feel more capable of participation in politics and more active members of local communities. These authors find that face-to-face deliberation has stronger effects on public attitudes than internet discussion. These micro results are strikingly similar. We now know that individual preferences, knowledge, and civic desires can be drivenat least in partby deliberation. But several issue remain outstanding. In almost all cases, it is unclear what aspect of deliberation is doing the causal work: justificatory quality, listening and respect, reciprocity and generalizability, or some other aspect. Although studies on preference and opinion change have begun opening this black box through the use of DQI, these results are provisional. Likewise, almost all of these studies come from experimental conditions. As such, it is hard to determine whether deliberation, face-to-face contact, moderation, information, the survey, or some other part of the process is generating the micro results we see here. Taking studies 'in to the field' will make it more difficult to control for some of these variables and therefore specify the scope conditions of deliberative benefits. But it would also help remove some potentially confounding variables (such as moderation), thus enabling adjudication between competing results (for an early attempt, see Jacquet, 2017). Moving forward, understanding which aspect of deliberation drive different resultsand under what conditionswill prove critical.

Meso-Changes: Social Learning, Polarization, and Consensus
A closely related body of literature has focused upon how deliberation influences groups or collectives. Much of this work also employs experimental techniques, though the methodological corpus is broadening over time.
A key part of this literature focuses on social learning: the ability of an actor to render their views understandable, denote the importance of that view, and make the reasons underpinning said view clear. 9 Proponents typically argue that inclusive and authentic deliberation will lead to deeper understanding and appreciation of the views of others. This concept is a close cousin of empathy, toleration, and recognition. 10 In early studies concerning the deliberative impact on empathy, Mendelberg (2002), in line with theories of motivated reasoning, suggested that linguistic intergroup biasthe privileging of arguments by members your own group at the expense of out-groupsis likely to undercut the ability of individuals to empathize. In a follow up piece, Mendelberg and Oleske (2000) found that discussion did not produce greater tolerance for opposing views nor mitigated conflict. This work suffered from a methodological bias, reliant upon participants' self-assessment for measures of empathy. 11 Recent studies, though, rely upon more objective measures. For instance, Siu (2008) shows that deliberation does shift opinions, but not in ways that reflect entrenched hierarchies, suggesting that deliberators are able to take seriously the views of others beyond stereotypical characterizations.
Other work in deliberative theory follows this more sanguine trend, though it is far from conclusive (Andersen & Hansen, 2007, 543). Michael Morrell (2010), in a pioneering study, contends that cognitive empathy (gaining knowledge on the other person's preferences and reasons for that position, juxtaposed against affective empathy which is knowledge of another person's mental state) is driven by reciprocity. Caluwaerts and Reuchamps (2014; see also Luskin, O'Flynn, Fishkin, & Russell, 2014) argue that deliberation acts as a buffer against more negative feelings towards the out-group, and demonstrate this finding in the context of two deliberative experiments in Belgium. These 9 Social learning is a 'group-level change' because it necessitates a dyadic relationship with another agent. 10 While related, social learning is not synonymous with recognition as the former focuses on understanding emotions or thoughts while the latter is concerned with appreciation of identity.
The results, however, should be of interest however to those concerned with identity politics. 11 On this critique, see Price and Cappella (2002). For instance, we should be concerned that individuals feel underappreciated even if their views have been understood and given due consideration by others.
scholars argue that the quality of deliberation does not produce more tolerance. Finally Grönlund, Herne, and Setälä (2015) have empirically demonstrated that deliberating in mixed groups over immigration (i.e. people both pro and con immigration policy) increases outgroup empathy. Deliberation in the mixed group, favoured by deliberative theory, generates higher levels of empathy than a con group also subjected to deliberative treatment. These results provide us with reason to think that deliberation has some effect, but the mechanism is unclear. This is especially complicated by the fact that deliberation is often said to require other-regardingness as a procedural good, making autocorrelation an issue. Separating deliberation into different features would make testing social learning as a dependent variable more theoretically robust.
Perhaps the most hotly-debated group effect of deliberation is whether the process leads to polarization (Farrar, Green, Green, Nickerson, & Shewfelt, 2009;Sunstein, 2002). In a series of publications, Cass Sunstein (2002Sunstein ( , 2009 argued that a 'law of group polarization' would develop from collective deliberation: groups that begin discussing an issue together will move toward the extreme of that position. This occurs through two mechanisms: social comparison and persuasive arguments. The former works as individuals seek acceptance of a group and see extreme positions as the safest 'bet' to adopt, while the latter operates as groups present arguments supporting the initial position, thus driving the group toward the predominantly held position (amplified by confirmation bias).
The literature on polarizationclosely linked to work on enclave deliberation has developed apace over the past decade. Myriad disciplines have produced results that group polarization does occur in many contexts. From law, Sunstein (2009) presents experimental evidence that groups tend to move toward the direction of the position initially dominating the group. In American politics, David Jones (2013) finds that workplaces dominated by partisan groups (especially Republican environments) shift toward that partisan extreme over time. In computer sciences and social psychology, Conover, Ratkiewicz, Francisco, Gonçalves, Menczer, and Flammini, (2011) use network clustering algorithms and manually-annotated data to study 250,000 tweets in the weeks leading up to the 2010 U.S. Congressional midterm. They show that Twitter users retweet information in ways that lead to polarization, but mention other users in ways that cross partisan divides.
Despite these (somewhat) robust findings, these studies are largely looking at group interactions or discussion. They do not focus on how deliberation impacts polarization. In these cases, results show the opposite: during deliberation groups come to learn about the views of others and thus depolarize (Grönlund et al., 2010). This is a result supported by Landemore and Mercier (2012) who argue that it is only groups of strictly-like minded individuals that fail to deliberate properly that are likely to polarize. Alternately, when conditions of reasoning are satisfied, deliberation guards against this outcome. Likewise Lazer et al. (2015) have shown that participation in a deliberative event induces individuals to discuss the topic with those who hold different opinions in society, thus undercutting polarization. Finally Lindell et al. (2017) have begun analysing not just group shifts in terms of polarization or moderation, but individual changes. These scholars find that ideological factors (a left-right orientation) are good predictors for the polarisation or the moderation of opinions. Interesting these authors find that the absence of immigrants in a deliberation about immigration seems to cause polarization toward the extreme, highlighting the importance of a politics of presence (descriptive representation). So while some scope conditioning factors are being explored, much more work is needed to determine whether clear arguments, reciprocity, or some other deliberative virtue helps guard against polarization.
Finally, early scholarship suggested that consensus was the goal of deliberation (Cohen, 1987). To the extent that deliberation generates single-peakedness we may expect that consensus results from deliberation. Over the years, however, this perceived benefit has been largely denounced by political theorists (especially difference democrats) and philosophers of language (Young, 2002). 12 It has also been suggested that consensus is only empirically plausible under highly stringent conditions, such as when deliberators share an underlying common interest and thick social bonds (such as friendship), as well as when the problem under discussion has an identifiable, "correct" solution (Delli Carpini, Cook, & Jacobs, 2004;Meldelberg & Okereke, 2002). Reasonable disagreement and heterogeneitywhich pervade almost all social contextsmakes consensus unobtainable.
Resultantly, there is now a prevalent notion that meta-consensus is a desirable outcome of deliberation (Dryzek & Niemeyer, 2006). Meta-consensus is the notion that post-deliberation agreement will emerge on the nature of the issue at hand. It is related tothough not synonymous withthe idea of intersubjective rationality in which deliberators come to agree on a decision for the same reasons. Employing data from a citizens' jury mini-public, Niemeyer and Dryzek (2007) have shown that deliberation does lead to meta-consensus, making decision-making more tractable. Providing related evidence, Grönlund et al. (2010) have shown that deliberators are able to construct common statements on complex topics after deliberation which, in turn, leads to higher rates of civic competence (i.e. willingness to participate more in future). This finding, they argue, is evidence that meta-consensus can emerge when groups need to work on constructing a joint position. In slightly more demanding way, Wesolowska (2007) shows that the structural conditions needed to reach meta- 12 The former because consensus risks silencing voices that do not conform to expectations of rational reason-giving, and the latter because of contestation over the nature of language.
consensus include previous common ground, positive evaluation of others' claims, and reciprocity.
There is now considerable evidence that deliberation produces otherregardingness and inhibits polarization. Work linking deliberation to (meta-)consensus lags behind. But across all three outcomes, there are very few studies showing what aspect (or combination of aspects) of deliberation matters most. As such, current studies are indicative of deliberation's potential, but not conclusive. In future, unpicking what element of deliberation is effectual and under what conditions requires a combination of experimental and nonexperimental studies. Qualitative case studies, for instance, will be important in understanding whether groups in public spheres can engage in social learning, inhibit polarization, or generate meta-consensus. While case studies focusing on deliberation's effect will be hard to undertake given the inability to control for non-deliberative factors (an issue of internal validity), it would greatly strengthen the external validity of current studies. Lab-in-the-field experiments, such as those run by Lazer, Sokhey, Neblo, Esterling, and Kennedy (2015), may well offer a useful 'half-way house' in this pursuit. This would allow scholars to test whether deliberative quality, and a wider array of latent variables, might matter in conditioning these outcomes in semi-controlled situations.

Macro-Changes: Popular Support, Deep Divisions, and Democratization
At its core, deliberative democracy is a theory of intrinsic legitimacy (Benhabib, 1996). Despite this, many scholars claim that deliberation will instrumentally induce individuals to support decisions even if they disagree with the outcome, help mend differences across views, and ultimately promote democratization as laws gain stability. These virtues operate at the polity-level: generating support in mass publics, overcoming societal cleavages, and democratizing shared political structures. 13 While experimental, qualitative, and quantitative research show promise on these fronts, results remain mixed. In line with micro and meso findings, there is some evidence that deliberation buttresses these outputs, but the precise deliberative mechanism underpinning said result remains understudied.
Also known as sociological or perceived legitimacy, many scholars have suggested that deliberation will generate popular support for a decision or rule (Manin, 1987). That is, citizens will supportor at least comply withmore deliberatively-justified decisions. This essentially Habermasian contention has been given credence by recent research showing that participation in decisionmaking drives perceived legitimacy of decision output. Peter Esaiasson and his colleagues, undertaking a randomized lab-in-the-field experiment with high school students in Sweden, find that personal involvement in decision-making is the strongest predictor of perceived legitimacy (Esaiasson Gilljam, & Persson, 2012). In another study, Denise Trabers (2013)employing interview techniquesfinds that participation in decision-making (especially when consensus is reached) generates the highest levels of perceived legitimacy for the decision. In one of the most direct tests of the linkage between deliberative democratic decision-making and perceived legitimacy of outcomes, Jenny de Fine Licht and her co-authors (2014) employ vignette experiments to test whether transparency in decision-making generates higher rates of perceived acceptance. The main finding from this study is that transparency in process and rationale enhances perceived legitimacy. In a boon for deliberative theory, these authors show that when decision-making approaches the deliberative ideal of respectful and rational argumentation, the highest rates of perceived legitimacy are attained.
These results, however, are tempered by a study from Mikael Persson and his co-authors (2013) where they show that democratic proceduresdirect voting and deliberationboth increase perceived legitimacy, though voting predicts for stronger effects. Additionally, Esaiasson, Persson, Gilljam, and Lindholm (forthcoming) report findings from twenty-seven field and vignette experiments on different subjects and comparative case analyses. These scholars show that outcome favourability -the extent to which a decision tracks an individual's preferencesis a stronger predictor of decision acceptance than procedural fairness. But given that deliberative democrats stress the endogeneity of preference transformation to decision-making, this is neutral finding. These studies do, though, give us good reason to think that some aspects of deliberation are more important than others in generating popular support.
Work on deeply divided societies has traditionally been agnostic or even oppositional to deliberation. For instance, some comparative democratization scholars have claimed that deliberation may stoke the embers of conflict (Horowitz, 1991). Alternately, consociational structures are supposed to foster bargaining between elites instead of involving citizens in violent agonism. Whether deliberation provokes or dampens conflict is an empirical question, however. Recent years have seen increased interest in how deliberation may help ethnically or otherwise divided societies overcome divisions (O'Flynn, 2007;Ugarriza & Caluwaerts, 2014).
Early comparative work focused on how elites in divided societies could be encouraged to deliberate instead of merely bargain. Steiner, Bächtiger, Spörndli, and Steenbergen (2004) showed that strong decision ruleseven in political systems characterized by deep disagreementcould generate deliberation. In a more normative vein, John  argued that divided societies should strive to decouple elite and public sphere deliberation to stop conflict over sovereign authority inhibiting dialogue between different groups.
Building upon this, many scholars have begun arguing that consociational structures should be designed with deliberative principles in mind so that individuals can deliberate across difference. For instance, Ian O'Flynn (2010; 2017) seeks to imbue consociationalism with deliberative norms to generate conflict resolution or, more strongly, to invoke shared intentions enabling societal divides to be bridged. These arguments, however, are mostly normative or conceptual. Direct evidence that deliberation in divided societies helps reduce conflict is hard to come by, and many scholars have turned toward the literatures on enclave deliberation and political trust for instruction.
Others, though, have exploited experiments in divided society for empirical data. Undertaking a deliberative poll in Omagh, Northern Ireland, Luskin et al. (2014) demonstrated that Catholics and Protestants were able to deliberate meaningfully, gain knowledge of opposing viewpoints, and ultimately support greater intermingling of ideas in policy outputs. Caluwaerts and Ugarriza (2014,), drawing upon nine case studies of divided societies from across the globe, argue that deliberation can lead to conflict resolution when institutions are decentralized (i.e. empower civil society) and elites promote engagement (i.e. lead by example). Steiner Jaramillo, Maia, and Mameli (2017), in a series of moderated deliberations between individuals from divided societies, find that rational argumentation and story-telling can sustain high levels of deliberative quality leading to agreement within the groups on key policy topics. Steiner's studywith a focus on how different modes and types of deliberation generate outcomesis indicative of the type of work needed to show not just that deliberation has desirable effects, but the mechanisms and contexts that are pivotal in these pursuits.
Work on national democratizationboth transition and consolidationhas been almost entirely insulated from discussions of deliberative democracy. As Dryzek (2009) notes, "the comparative study of democratization has missed what, to many analysts and democratic innovators, is the most important aspect of democracy: deliberation" (p. 1379). There is, however, an emerging literature that shows tentatively positive results linking deliberation in civil society to national democratization.
Importantly, a major strand of empirical deliberative analysis has focused on formal state institutions critical for flourishing democracies (such as legislatures and courts). Steiner et al. (2004), using the DQI to analyse 4,488 speeches from German, Swiss, and U.S. parliamentary debates, find that deliberation can develop within these institutions. However, there is variation across these bodies: Swiss grand coalitions enhance respectful behaviour of MPs much more than the US Congressional rules and German Parliamentary procedures. In a more negative vein, Landwehr and Holzinger (2010) compare a German parliamentary debate to a citizen conference on embryonic stem cells and find that Parliamentary debate is less deliberative than the alternative. These studies, though insightful, do not show a link between deliberation in formal bodies and the ability of a polity to sustain democratization.
Some work attempts to link deliberative quality in society to the transition toward, and consolidation of, democratization. The clearest evidence linking deliberative quality in formal and informal spaces of the state to democratization comes from the literatures on constitutional moments. Jon Elster (1998), analysing the French and American experience, depicted an hourglass model of how public deliberation should shape constitutionalism. He argued that democratic stability would be best ascertained through public participation in initial stages with public ratification at the end, while the writing itself should be shielded from citizen input. This model has recently been shown limited. In the most comprehensive study to date linking citizen deliberation and national democratization, Eisenstadt, LeVan, and Maboudi (2015) analyse 138 new constitutions in 118 countries between 1974 and 2011. By disaggregating constitution making into three stagesdrafting, debating, and ratificationthese authors demonstrate that public input early in the drafting stage significantly increases the stability and democratic quality of the regime over time. Seeking support later in the process is much harder to achieve. These authors claim that this lends significant support to deliberative democratic theory, but do not directly study deliberation itself in these moments, instead taking inclusiveness, transparency and participation of citizens as proxies.
These polity-level studies have many limitations and therefore future directions for research abound. Research on public support should move beyond experiments and draw on studies of citizen behaviour in democracies, which seek to understand whether individuals are more accepting of policy decisions when elites justify their positions deliberatively (Colombo, forthcoming). Studies on deep divisions have made good use of experimental minipublics in some contexts, but looking at truth and reconciliation commissions on the ground is a useful avenue. Likewise, time-series data that tracks elite deliberative quality vis-à-vis societal cleavages would be instructive. In terms of democratization, there are very few studies linking deliberative quality in formal institutions and civil society to national democratization. Most large-n datasets on national democratic quality -Freedom House or Polity IVdo not contain coding on deliberative indicators. This is changing, though: V-Dem, a new comparative and longitudinal dataset on national democratization, contains a range of deliberative indices. This will make it possible for future scholars to take up the task of examining if, how, and under what conditions deliberative quality in different parts of the state promote transition and consolidation of democracy (but see Coppedge, Gerring, Altman, Bernhard, Fish, Hicken, & Semetko, 2011). Overall, these studies need to assess how the supposed instrumental benefits at the macro-level are related to different aspects of deliberation between citizens and elites, while also focusing on the scope conditions under which these benefits accrue.

The Priority of the System: Tying the Levels Together
The key finding from the previous section is that deliberation does generate a swathe of beneficial outcomes, but there is variation across these effects (for a summary, see Table 1 above). The main frontier for research, already underway, is determining which aspect of deliberation does the causal work in generating different benefits. For instance, more work is required to isolate whether respect, high quality justifications (i.e. with multiple reasons connected to arguments), reciprocal or generalizable arguments, willingness to compromise, or any other featureor combination thereofmatters most. Precisely because different instrumental effects may beand likely aregenerated by different mechanisms, this work will provide critical. This work is especially relevant as scholars are increasingly taking a 'functional' or 'problem-driven' approach to deliberation (see respectively : Bächtiger & Beste, 2017;Warren, 2017). These theoretical moves recognize, in one way or another, that deliberation is actually a composite concept comprised of many different mechanisms. This 'nonunitary' view holds that the 'right' deliberative or otherwise democratic action to be invoked depends on the issue at hand: we should tailor prescription to the problem needing solving. In some instances we need to overcome deep divisions, and other times knowledge gain is crucial: as different aspects of deliberation may generate these results, we need to disaggregate deliberation and link mechanisms to outputs. Moreover, as some features of deliberation may work well in tandem, it is critical to begin this disaggregation process and see how certain features, in isolation and in tandem, are linked to different outcomes. While certainly complicated work, the theoretical and empirical importance cannot be understated.
Relatedly, with some notable exceptions from micro-studies, the scope conditions that inhibit or enable deliberation have not been studied in depth. This includes thinking about how non-deliberative elements of the experimental approachsuch as face-to-face interaction with deliberators, moderation, and information provisionalter outcomes. Likewise, the psychological makeup of participants and their backgrounds (class, gender, race, etc.) may all matter in the outcome produced. Finally, how deliberative events are embedded in social and political contexts will also likely impact whether certain benefits are achieved. Case studies and large-n analysis will raise the stakes in terms of the number of variables that bias results, but are essential in thinking about how deliberation operates in the 'real world'. 14 However, understanding when and why to pursue these instrumental benefits requires deeper engagement with normative theory. Fortunately, the two pathways forward outlined in this article are both intertwined with the recent systemic turn in deliberative theory. Time is therefore ripe to assess how these instrumental effects may impact systemic theorizing and vice-versa. I undertake this task in three steps. First, I reflect upon the core normative goal of the systemic turn and situate my position against others in the literature. Second, I discuss how deliberative systems can and should make use of the empirical evidence marshalled in this article to combine micro, meso, and macro levels in this pursuit. Finally, I show how key implications of the systemic turn should inform future analysis of the instrumental benefits of deliberation at the three levels specified here.

The Systemic Turn
The basic premise behind the systemic turn is that deliberative democracy cannot be limited to formal reason-giving in hermetically-sealed fora. Instead deliberation should be scaled up: distributed across interconnected institutions in time and space. Differential institutions will play different roles in this systemic architecture, and individuals will be able to contribute arguments and reasons through a range of communicative modes in those venues. The systemic view, then, strives to combine the micro-element of individual deliberation with meso-importance of discreet fora and the macro necessity of polity-wide structures.
This systemic turn has proven to be a fruitful line of inquiry. Its roots hark back (at least) to Habermas' (1996) two-track notion of democratic societal legitimacy in which decision-making in empowered spaces is directed by deliberation within and between publics. And we now have different ways of assessing the democratic legitimacy of deliberative systems. In roughly chronological order, Robert Goodin (2005) has outlined a distributed model of deliberation in which different institutions play different roles in discharging the deliberative ideal. Carolyn Hendriks (2006) provides an integrated model of deliberative systems in which micro, macro, and mixed discursive spheres enable different representative and participatory outlets for citizens. Dryzek (2009) focuses on whether a system enables inclusive, authentic, and consequential deliberation as a metric of (democratic) legitimacy. Parkinson and Mansbridge (2012) look at how institutions in a systemic division of labour can be coupled together in ways that enable the flow of good arguments but inhibit co-optation (see also Mansbridge, Bohman, Chambers, Christiano, Fung, Parkinson, Thompson, & Warren, 2012). Their undergirding commitment is to provide a system that enables ethical, epistemic, and democratic values to emerge. For a summary of these positions, see Table 2. Division-of-labour emerges between the parts in generating epistemic, ethical, and democratic goodness; coupling between institutions should allow flow of arguments but restrict co-optation These systemic views have emerged largely in response to criticisms raised against deliberative democracy during the 1990s and 2000s. Difference democrats, for example, accused deliberative scholars of focusing too heavily on justificatory rationale, and ignoring other modes of reason-giving (Young, 2002) while agonists argued that a search for consensus robbed politics of its contestatory nature (Sanders, 1997). More deliberative-friendly researchers suggested that the focus on isolated fora (such as minipublics) shifted attention too far away from mass publics (Chambers, 2003), or that minipublics cannot legitimately make public policy (Lafont, 2015;Parkinson, 2004). 15 The systemic turn tackles these criticisms directly by asking how the deliberative ideal of reason-giving might be compatible with different forms of communication, and what the appropriate space for non-or even antideliberative action (protests, violence, etc.) might be (Fung, 2005). This turn also marks a conscious return to thinking about how mass publics can be knitted to binding decision making and has gravitated toward understanding how minipublics might be differentially utilized in broader systemic contexts for systemic legitimacy (Goodin & Dryzek, 2005;MacKenzie & Warren, 2012).
Despite these advancements, the systemic turn has also come under heavy criticism. The major reproach is that adapting deliberative democracy to empirical 'facts' risks cutting loose the normative anchor of deliberation. This argument comes in two forms (Owen & Smith, 2015). First, there is a worry that the breakdown of the unitary modeland evaluations of legitimacy and justice at the systemic levelmeans that deliberative systems may actually entail very little deliberation. By evaluating the system as a whole, distributed instances of deliberation might not ever mean that the ideal of deliberative democracy is reached, and citizens do not 'reason together'. Second, the admittance of novel forms of communication and the acceptability of nondeliberative acts have been interpreted as an abrogation of the normative core of deliberative theory for which clear bounds of acceptability have not been specified (Bächtiger & Parkinson, forthcoming).
These are importance considerations. Constructively, Owen and Smith (2015) stipulate two alternate ways to advance the systemic turn overcoming these issues, while (allegedly) being compatible with the ideal of deliberative democracy. There first suggestion is that systems should embed a deliberative stance: individualswherever possibleshould take up the ideal of deliberation and enact a form of collective practical reasoning that undergirds collective decision-making. The second notion is that we should move toward a democratic system: here deliberation is just one mechanisms that legitimates collective decisions alongside voting, exit, contestation, etc. (see also Warren, 2017).
Both suggestions are problematic. The notion of a deliberative stance misunderstands the normatively value of deliberation. Precisely because we have empirical evidence that some deliberative features are better suited to bringing about some outcomes, a breakdown in the unitary model is both normatively and empirically compelling. Owen and Smith (2015) might employ a reductio ad absurdum in which a deliberative system has no 'real' deliberation. But this point again misrepresents the nature of the deliberation: it was always a regulative ideal that is only able to be approached in social and political contexts. Working out how to approximate the ideal has always been the central task of deliberative democracy. This opens space for thinking about when a breakdown of the unitary model may be justified. The claim relating to democratic systems should also be rejected. Deliberative democrats qua deliberative democrats should hold on to the position that legitimacy is attained through the rational argumentation between affected individuals. Nondeliberative modes of collective action can only be rendered acceptable through justification, thus giving deliberation primacy in the design, operation, and outcome of collective decision-making (Dryzek, 2009;Thompson, 2008). While it is important to specify how these meta-theoretical justifications should occur, this is a matter of argumentation, and not a reason to abandon deliberation as the motor of the systemic legitimacy.
As such, deliberative democrats require an alternate way to justify the systemic move. I suggest the following: deliberative systems ground ethical, epistemic, and democratically legitimate decisions to the extent that actors and institutions are arranged to ensure that 'better arguments' win out in the process of collective decision-making. The notion that arguments which are expressed honestly, with factual and supporting reasons, while attaching to shared norms stands at the core of deliberative theory. 16 The process of giving and taking arguments treats others as an ethical source of reason, sorts good epistemic claims from bad ones, and generates legitimacy for collective decisions through the inclusion of affected parties. This does not necessarily mean that there is always one 'best argument' for any given issue. Rather, that through the process of deliberation, weaker arguments are separated from better ones, and new argumentsperhaps better than those held by any individual, group, or faction can be made manifest through different forms of collaborative, creative, and contestatory argumentation. Institutional structures in a deliberative system should be arranged to enable better arguments to be formed in different parts of the system, with binding decisions reflecting the weight of better argumentation. 16 There is not space in this piece to discuss, at length, what constitutes a 'better argument' as there are post-modern, psychological, and post-colonial debates which question the nature of empirical truth, truthfulness, and moral rightness. However, I follow Habermas (1996) classic formulation here. In defense, I simply note that, insofar as deliberative democrats think better arguments can win out in the ideal speech situation, there must be better or worse forms or argument to be uncovered which are related to truth, sincerity, and moral norms.

From the Individual to the Group and on to the Polity
By adopting this view of deliberative democracy, it helps make sense of the role of micro, meso, and macro deliberation. My argument is that pursuing instrumental benefits at these levels should be contingent upon whether it foments better arguments across the system, ultimately transposing those arguments into binding rules. This argument tracks other claims in the literature that the systemic level has priority in normative analysis, but with more specificity. By this, I mean that the micro-, meso-, and macro-level do not necessarily have priority over one another in the evaluation of how a system discharges its core goal. This might seem counterintuitive, but the priority is non-obvious. For instance, we should not sacrifice national democratization or popular support for citizens with a lack of knowledge or minimal civic desire (lest they elect poor leaders or do not participate at all). Similarly, we should accept polarized groups that craft well-founded arguments instead of blindly shifting opinions (in line with social pressure) or deferring to difference across divided societies (i.e. accepting the views of others with no good reason). As such, there is a clear systemic priority: the justifiability and design of deliberative systems must allow better arguments to emerge and eventually win out in the decisions that bind all those within the system. So how exactly does this normative aim for a deliberative systems help guide institutional design? Following Habermas (1996) and Dryzek (2009), the distinction between empowered and public spaces is crucial. 17 The basic division of labour between these two spheres should be as follows: publics spheres are arranged to allow the strongest arguments on matters of common interest to be formed. These arguments should be then transmitted to the empowered space, whereby these arguments are trialled against one-another on equal and non-coercive terms. This trialling of argumentation in the empowered space then pits the best possible arguments against one-another. To give this framework more specificity, it is important to be clear about what normative distinctions turn upon each space, how actors in each space should thus operate, and how this relates to micro, meso, and macro deliberation.
In public spaces, actors should be provided space to formulate arguments in the strongest possible terms. Precisely because public spheres contain a range of imbalances (power, epistemic, etc.) this should be considered in justifying institutional arrangements. Whether institutions should be coupled together or shielded from wider societal processes depends on how better arguments will be formed. If groups need to test their arguments against other claims in the system to strengthen their view, then interlinkage is desirable. This highlights the importance of systemic thinking, as deliberation may often require individuals in isolated groups to be open to collaborative engagement with actors in other parts of the system in order to form new and better arguments. Alternately, if a group may be co-opted or colonized through such linkages, then institutional dislocation is likely required for the formation of better arguments, especially as actors seek to overcome historical or structural power imbalances emanating from other parts of the system. Through these differentiated processes, different publics arrive at their best arguments.
In the empowered space, however, such differentiations are not normatively justified. Rather, we should rather demand formal equality of reason-giving in shaping law and policy. 18 As such, while some steps (i.e. quotas) might be necessary to achieve formal equality, what matters is that all arguments are tested against each other in crafting binding rules. While there are many times in political life where disagreement will still exist after better arguments have removed some (or even many) options, this is when different forms of deliberative bargaining should pervade (Mansbridge, Bohman, Chambers, Estlund, Føllesdal, Fung, & Martí, 2010). By trialling the best arguments from different parts of the public space against one another in the empowered space under conditions of equality, this process allows individuals to see that their views are taken seriously (publicity), provides critical distance from the state to allow free public opinion-formation, and enables a form of 'reasoning together'.
This analysis leads to the claim that pursuing micro, meso, and macro instrumental benefits is conditional upon whether it helps promote the better argument at the systemic level. This is a tricky guideline to follow in practice, but it does mandate certain actions. For instance, in many cases we will think that individuals developing better knowledge or enhancing social learning is desirable. However, while we should demand this of actors in empowered spaces who need to learn about the views of others to appropriate subject their own claims to reasoned argument, this might not be desirable for individuals or groups in public spaces who should focus more on honing their specific claim rather than learning about a range of alternatives. Similarly, while polarization is usually considered normatively suspect, 19 it is desirable when it helps groups overcome systemic pressures that dampen the creation of their arguments. Finally, we should not simply accept macro democratic systems that allow micro or meso deliberation to feed in to empowered spaces without asking whether these are the best arguments for grounding systemic legitimacy. If seeking micro or meso benefits are actually removing better arguments from the system, then we should question the justifiability of those instrumental goals. If two instrumental benefits stand in tension (say, between knowledge gain and meta-consensus during group deliberation), then working out how to make the trade-off should be done in accordance with the systemic goal outlined here. Similarly, thinking about the right constitutional design or democratic structure (i.e. parliamentary vs. presidential) at the polity levels requires thinking about how better arguments can be made across the system.
It is for these reasons that more knowledge about the drivers of instrumental benefits and their scope conditions is so important in systemic theorizing. By understanding how these instrumental effects work, we will be better able to prescribe how and when to induce certain outcomes and when to negate them. For instance, if a group needs to polarize to produce a strong argument, knowing how polarization is stymied is essential in avoiding this prescription. In this sense, empirical analysis is crucial in thinking about the applicability of normative principles.
Ultimately, public spaces should be arranged to allow better arguments to surface. As this occurs, empowered spaces should weigh these arguments against one another under conditions of equality as binding rules are decided. It recognizes there is no priority between micro, meso, and macro levels, but rather sees each level as important insofar as they contribute to a system. But it does have implications for how we justify institutional arrangements differentially in empowered and public spaces. Pace Goodin and many others, it means that we should limit distributed deliberation to public spaces rather than empowered spaces because, in the end, empowered spaces must trial the best systemic arguments against one another to ensure better arguments win out. While distributed deliberation in public spaces might be useful in forming better arguments, it does risk segmenting arguments. Distributed deliberation in empowered spaces could only be justified, then, if all relevant arguments are tested against one-another at some point. Similarly, while polarization in public spaces might be good for crafting stronger arguments, it is likely problematic in empowered spaces as it would entail ignoring key arguments. In this way, my claim gives priority to the systemic level, but employs micro, meso, and macro deliberation in different ways depending on whether it occurs in public or empowered space.

Novel Forms of Communication and Non-Deliberative Acts
Although we need to think about how micro changes, meso alterations, and macro outcomes sustain a deliberative system, we should also think about how the systemic turn can impact analysis at those levels. Two issues stand out. The first is to begin using new modes of communication to think about how instrumental values might be attained. The second is to start using the scope conditions noted above to think about the design of a system in holistic terms. This will be especially important for the justifiability of non-deliberative acts. I discuss both in order.
A major reason behind the systemic turn is that the Habermasian emphasis on formal reason-giving elided the importance of alternative form of argumentation. It is now widely agreed that narrative, rhetoric, story-telling, humor, art, and many other modes of expression are legitimate forms of argumentation. While normative theory has rightly advanced this claim, and scholars have begun thinking about how different communicative modes are used across a system, there has been very little feedback to those working on the instrumental benefits of deliberation. For instance, I know of almost no studies that attempt to systematically study how different forms of reasongiving impact micro, meso, or macro deliberation. 20 It is likely that preference/opinion change, knowledge gain, and civic participatory desires are driven by different forms of argument depending upon the individual. Likewise, meso and macro changes are almost certainly driven by different types of communication depending upon how those modes manage to tap in to, or disrupt, group norms. For instance, empathy is often established by finding new ways to understand another's position, and popular support seems very likely driven by how elites 'perform'.
As such, it is imperative to take these new forms of communication and see whether certain micro, meso, and macro level benefits are enhanced or impeded by their use. This, however, is an empirical issue. And, while establishing how different forms of argumentation matter for instrumental benefits, their usage should still be determined on normative grounds: whether better arguments are produced across a system. Of course, this normative goal should make room for the fact that alternate forms of argumentation is likely intrinsically valuable as self-expression is a key to autonomy. But it seems plausible to me that alternate forms of argumentation are also crucial to the formation of better arguments as individuals can express themselves in ways that attach to their cultural, social, and psychological make-up. When we understand how micro, meso, and macro benefits can be achieved, then we can begin making systemic trade-offs between those benefits and ensuring that better arguments are produced in different institutional locations.
A similar point can be made about the utilization and justifiability of nondeliberative acts. While most scholars have accepted the expansion of communicative acts, the import of non-deliberative acts (such as protests and violence) has been heavily criticised. Though it is very plausible that such acts might eventually generate systemic benefits, a key problem remains: there is a very high epistemic threshold in understanding when non-deliberative acts have good outcomes (Fung, 2005). The social world is complex, maybe even stochastic. If we do make space for good deliberative outcomes based on nondeliberative actions, then we need strong knowledge of the determinants and outcomes of different deliberative situations. For instance, it is an empirical argument whether non-deliberative situations (i.e. protests) enhance or undermine national democratization over time. To the extent we favour this outcome, the defence of protests over government control should be informed by relevant empirical evidence.
Understanding what benefits micro, meso, and macro benefits bring is important in (at least) three ways, then. First, it helps comprehend said benefits even under circumstances of social complexity and thus enables the construction of more reliable prescriptions. Second, this information is useful insofar asby telling us what deliberation producesit will also tell us what deliberation cannot produce. Perhaps deliberation is good at reducing polarization when groups are composed of different views and a politics of presence is available, but deliberation is not be able to undo the effects of polarization (conceivably due to motivated reasoning) when such conditions are unavailable. When such conditions do not obtain, then, we have strong reasons to justify nondeliberative remedies. Righting systemic wrongs through non-deliberation can then be informed through empirical knowledge of what deliberation cannot accomplish in the production of better arguments.
Finally, by thinking about the use of non-deliberative acts, it also turns our attention toward the scope conditions of deliberation. Section II argued that more work is needed to think about the conditions under which instrumental benefits are attained at micro, meso, and macro levels. This includes issues such as the role of information and moderation in mini-publics, the individual characteristics of deliberators, the type of political system in which institutions are embedded, and so on. By thinking about these factors, we will know how to tie institutions into a system in ways that promote better arguments. Do some participant characteristics impede instrumental benefits? Do some representative systems (first-past-the-post vs. party list) alter the demands of government justifiability and popular support/democratization? Does linking a minipublic to a parliamentary body limit or promote civic participatory desire? Do non-deliberative acts help or hinder recognition across difference in other parts of the system? Thinking about these question is important. But crucially, thinking about these questions and asking whether it enhances or undermines the quality of arguments in a system should remain the final consideration.

Concluding Remarks
In conclusion, deliberative democracy is now a normative theory with established empirical implications. We know much more about the types of beneficial outcomes produced by deliberation across different contexts. However, in several ways, this literature remains embryonic. Few studies seek to disaggregate deliberation in to its relevant features and assess which aspects are most important for obtaining different values. While studies focusing on preference shifts, opinion change, knowledge gain, and (to some extent) social learning and popular support look at the quality of deliberation as a driver, the link between other beneficial outcomes and deliberative quality remains understudied. The scope conditions that produce these benefitsand how this gleaned evidence can help think about systemic designshas only just begun. 21 In terms of future work, a large portion of evidence comes today from experimental (lab, survey, and field) designs. While the usage and utility of these methods should not be understated, scholars should continue to reflect upon issues of internal validity (whether the effects of deliberation can be separated from other parts of the process) and external validity (problems of self-selectionwhether those who opt in to deliberative exercises vary in some significant way from the wider population they are supposed to represent). 22 In order to mitigate these issues, qualitative scholars and practitioners, both separately and in collaboration, should seek to test the effects of deliberation in the field (improving external validity). Quantitative scholars should also seek to expand the range of comparative cases available to study and improve the operationalization of deliberation to enhance both internal and external validity. Likewise, several other instrumental benefits should be considered: social trust, emancipation, distributive justice, and epistemic correctness have all been purported, though studies remain scarce.
Ultimately, though, valuing deliberation for its instrumental effects is still a normative commitment (Thompson, 2008). While our intuitions can only be defeated by other intuitions, our reasons for holding these views can be supported by empirical evidence. In addition to whatever intrinsic reasons one has for being a deliberative democrat, then, we also have (some) good instrumental reasons to support this position. I have suggested that the ultimate systemic test should be whether the institutional landscape enables the emergence of better arguments in public spaces to be equitably trialled against one-another in the empowered space. This way, better arguments form binding decisions. This has implications for how we justify different institutional arrangements (such as distributed deliberation), the importance of new modes of communication, and the use of non-deliberative acts. But most directly, it suggests that the instrumental values of deliberation should be understood in terms of how systems produce turn better arguments in collective decisions. When symbiosis occurs between micro, meso, and macro deliberation and this systemic goal, then we should promote the instrumental aims. Where trade-offs emerge, however, the systemic goal must take priority.