A relational perspective on deliberative systems: combining interpretive and structural analysis

Deliberative systems theory suggests that a democratic society works well when citizens’ experiences and views, as expressed in various forms and sites of communication, are connected and taken up by other citizens as well as policy-makers. Pluralism, which is not always easily reconciled with high-quality deliberation in every instance, is seen as instrumental to the realization of democratic values and sound decision-making. This perspective raises new methodological challenges, such as (1) identifying sites of communication that serve important functions in a deliberative system, (2) connecting different sites and (3) assessing their impact. Recent scholarship has found that these challenges can be fruitfully met by applying interpretive methodology, which, like deliberative systems theory, aims to understand social interactions on their own terms, and not by measuring their correspondence to theoretical ideal-types. However, for theory development, as well as to help improve actual deliberative systems, researchers also need to make generalizable inferences. This paper develops a relational approach that combines interpretive methods with structural theory, which allows researchers to assess and explain the deficiencies, as well as the opportunities, that citizens experience. The principles of relational analysis are illustrated by research on citizen deliberation about urban riots.

The current development of a systemic perspective on deliberation, which sees communicative sites as parts of a network of discourses, provides a crucial democratic perspective on governance (Mansbridge et al. 2012;Stevenson and Dryzek 2014). This analytical perspective, which is still much under development, has generated fruitful new discussions about how societies can become more democratic and egalitarian, and how public policy and implementation processes can gain deeper legitimacy and effectiveness. Its central concern is how inclusion and uptake of a plurality of perspectives and types of knowledge affect the capacities of organizations and societies to handle new challenges. In contrast to previous perspectives on deliberation, which typically focused on particular types and models of public reasoning as examples of deliberative democracy, the systemic view acknowledges that societies depend on knowledge and ideas that develop through various types and sites of communication (Elstub, Ercan, and Mendonça 2016;Mansbridge et al. 2012;Dryzek 2009). Beyond the theoretical debates it has generated, it promises to increase the capacities of publics to accurately assess democratic deficits.
This new development raises important methodological questions for empirical research, since it is not obvious which acts, events and forums may actually play significant roles in a deliberative system (Stevenson and Dryzek 2014;Ercan, Hendriks, and Boswell 2017;Elstub, Ercan, and Mendonça 2016;Moore 2016;Smith 2016). Which sites matter most to citizens and to democratic institutions? How are different sites connected? How do they make an impact? These are important questions for assessing a system's capacity to create conditions for pluralistic deliberation and effective uptake. According to recent contributions, they are best answered through interpretive analysis, that is, approaches that focus on subjective meaning and experience in relation to language use, acts and interactions (Yanow 2003, 11). Like deliberative systems theory, interpretive research aims to understand social interactions on their own terms, rather than measuring their correspondence with theoretically derived ideal types (see Ercan, Hendriks, and Boswell 2017; see also Mendonça 2016;Hendriks 2016;Boswell, Hendriks, and Ercan 2016;Blue 2016).
Interpretive methodology has its limits, however, which are due to its dualistic understanding of the relations between subjective and objective and understanding and explaining (Bevir 2011b;Yanow 2003). As I will argue, several important contributions to research on deliberative systems have successfully bridged the theoretical divide between structural analysis and interpretive analysis, but this has not been reflected in methodological discussions. In such discussions, it is still claimed that such bridging goes against the principles of interpretive research. I argue that distancing oneself from 'objective knowledge' and from explanations of social action that focus on interests and incentives rather than meanings and experience, makes it more difficult to address some questions of inclusion and uptake in deliberative systems, such as, how specific uptake problems are related to more general weaknesses in a deliberative system. The methodological considerations therefore have concrete and important consequences for our ability to address democratic problems and find ways to improve deliberative systems. For example, how do meaningful exchanges between citizens depend on effective social institutions? How do the ways issues are framed in political discourse depend on social and political structures? And how are particular modes of communication affected by organizational opportunities to voice concerns and have one's perspective considered by others through respectful and open-minded interaction? These are questions that concern the interactions between structural and institutional conditions and subjective meaning and motivation. They concern understanding subjective meaning and explaining action in terms of incentives and interests.
Answering such questions, requires a combination of interpretive and structural analysis, or what Emirbayer (1997Emirbayer ( , 2010) calls a 'relational' approach. A relational approach neither focuses on meaning as such, nor exclusively on structures, but on the relations between structures and social interaction. In a relational approach, interpretive methods facilitate deep learning of social meaning and practice, while structural analysis provides tools for understanding constraints and possibilities for inclusion and uptake. Central questions concern the relations between the subjective, or interactional, and structural, or objective, aspects of a situation. It asks: how do structures affect/ reflect social interactions? And how do social interaction affect/reflect structures? For democratic governance, such questions may concretely concern, for example, how structurally affected prejudices against ethnic groups, and structurally conditioned status differences, influence the uptake of knowledge and experiences of participants with ethnic-minority backgrounds. Conversely, they also concern how such interactions affect the possibilities of social change (Hayward 2000;Hayward 2013;Author 2016). It is therefore the claim of this paper that relational analysis can play a crucial role in eliminating obstacles and distortions that undermine the potential of deliberation as a means to provide insights and viewpoints to inform decision-making and governance practices.
The paper begins by discussing the arguments for applying interpretive methods in studies of deliberative systems and the limits of such methods. In the second section, I describe what a relational perspective on deliberative systems would mean. In the third section, I discuss notable examples of research on deliberation that have been described as examples of interpretive analysis. I argue, that their important contributions to deliberative theory are due, not only to careful interpretive work, but as much to their (implicit) application of relational analysis. To learn from these examples, it is important to acknowledge the structural component. In the fourth section, I outline three general steps of relational analysis of deliberative systems (decentering discourse, identifying structural circumstances and reconstructing mechanisms of change). In the fifth and final section, I illustrate these steps by describing a case study of deliberation after the occurrence of riots the city of Stockholm in 2013. As the case shows, relational analysis has the potential, not only to problematize exclusions and uptake failures, but also to contribute constructively to improvements of a deliberative system.

Interpretivism, deliberative systems and structural analysis
The central characteristic of interpretive methodology is its focus on subjective meaning as an area of research, and its emphasis on understanding rather than explaining, as its research motive. Interpretive research explores the subjective experience of empirical phenomena rather than seeking to identify specific objective structures and causal relationships (Schwartz-Shea and Yanow 2013;Yanow 1999Yanow , 2003. This approach to research has clear affinities with deliberative theory, and the deliberative turn further accentuates them. Interpretivism and the systemic approach to deliberation share the idea that subjective experience is epistemically valuable independently of objective validity, and both aim to 'decenter' the exploration of social interaction, by resisting to measure its relation to objective standards, for example by applying a normative theory of deliberative democracy to each communicative setting (Owen and Smith 2015;Ercan, Hendriks, and Boswell 2017). Applying interpretive methodology, means 'a decentering of expertise on the part of the researcher: accessing local knowledge of local conditions shifts the researcher's expert role from technical-rational subjectmatter expertise to process-expertise' (Yanow 2003, 11). Hence, the interpretive approach implies learning through the social meaning of interaction, rather than its objective meaning, which is, writes Yanow (2003, 11), 'a radically democratic move.' Interpretivism means thinking of research as understanding together, a process of making sense of a phenomenon by building on local knowledge and perspectives. It shares with deliberative theory an emphasis on empathy as a condition of understanding (Yanow 2003, 11;Mansbridge 1983, 27).
The 'systemic turn' meant that the connection between deliberative theory and interpretive methods was further accentuated. While empirical research on deliberation has grown fast during the last two decades through case study research and experiments, the methods used did not usually reflect the theoretical development that occurred simultaneously (Ercan, Hendriks, and Boswell 2017). Often focused on isolated cases and experiments, they often missed connecting the resultant interactions to the wider political and societal context (Mansbridge et al. 2012). Moreover, rather than measuring the outcome in terms of 'deliberativeness,' the systemic perspective suggests that various forms and kinds of communication may impact significantly, and independently of 'deliberative quality', on the overall deliberative capacity of a whole society, or the system taken as a whole (Stevenson and Dryzek 2014, 32). The parts of the system should therefore be evaluated more on their own terms, that is, as sites that may give important contributions to public deliberation in ways other than by fulfilling criteria of good deliberation. For example, partisan protest may often provoke critical reflection and deliberation more effectively than nuanced reasoning. Moreover, 'everyday talk' between citizens may often not involve the kind of give and take of reasons and arguments that ideal deliberation would include, but have properties that make it socially meaningful and strengthen the bonds between members of a community, which improves conditions for deliberation (Mansbridge 1999). Interpretive methods offer the kind of flexibility and sensitivity to context and experience that is crucial to uncover the values of such 'non-deliberative' events and everyday interactions for a deliberative system. Interpretive analysis may thereby contribute significantly, and in ways that other types of analysis cannot, to the knowledge needed to create governance practices that are more effective in its uptake of different kinds of knowledge and insights.
The difference that interpretative methods makes in deliberative research becomes especially clear when seen in relation to analyses informed by a structural perspective, that is, a perspective which focuses on the incentives provided by societal and economic conditions and how they relate to different actors' interests (Elder-Vass 2010). When employed in analyses of deliberation, such perspective leads to a reductive understanding of deliberation, since it assumes that participants will be motivated by self-interest (e.g. Przeworski 1998;Shapiro 1999;Rodgers 2012). Structural analyses do not, on their own, acknowledge the possibilities of other motives, nor of an interaction between structural conditions and deliberation in which the latter might also affect the first, and not just the other way around (Emirbayer 2010). Structural analysis may clarify the obstacles and opportunities that actors face, but typically neglects to account for the role of actors in modifying such obstacles and opportunities, or reproducing them. It therefore downplays, for example, the role of deliberation as a generator of new knowledge and critical reflection, which can effect change (see also Bohman 1997).
Structural analyses and interpretive methods are as conceptually at odds as theoretical perspectives as are deduction and induction as research strategies. Combining them requires moving between structure and agency, institutions and identity and objective conditions and subjective meaning (similarly to how 'abductive' research moves between deductive theory-application and inductive fieldwork; see Hay 2002).
Without structural analysis, interpretive methods are limited to interpreting the meanings of language use, interactions, objects and events. This is an important limitation in studies of deliberation, not the least, because it makes it difficult to answer some central questions concerned with power, such as how a site of deliberation may be constrained by institutional conditions, or how it may impact on wider social structures. It also makes it difficult to examine actors' interests and how they affect interactions between people depending on social positions, economic incentives and political conditions. These questions are crucial for assessing, and improving, the way deliberative systems work. Hence, similarly to structural analysis, interpretive approaches do not completely answer the crucial questions raised by deliberative systems theory. To answer them, it is necessary to combine structural and interpretive analysis. As I will argue, several important examples of interpretive research on deliberation has done this.

A relational perspective on deliberative systems
Interpretive methodology is commonly seen as belonging to a different tradition than research that aims for theoretical inference, explanation and generalization (see Yanow 2003;Ercan, Hendriks, and Boswell 2017, 198). However, interpretive approaches are not incompatible with the ambition to capture phenomena using structural analysis. On the contrary, an exploration of subjective meaning can often be more insightful if it also accounts for the structural conditions that facilitated that understanding, including economic conditions (see Calhoun 1991). Take, for example, the way public deliberation has come to be understood as a social institution that has taken a specific form through interaction with economic and social structures (Habermas 1989;Fraser 1990;Calhoun 1993). This understanding of the development of public deliberation in Western democracies also has important implications for discussions of deliberative democracy today, especially with regard to gender equality and multiculturalism (Fraser 1990), since the question of how structures affect the possibility of participation on equal terms continues to be central for deliberative theory.
Relational analysis is one way to combine interpretive methods with a structural perspective. Other options may be divided into two broad types. The first interprets the datause of language, acts and interactions or objectsas situated in, and produced by, structural conditions. From this perspective, the reason that interpretive methods should be combined with a structural perspective is that social meaning is significantly conditioned by economic and social structures (Calhoun 1991), but social meaning is not in turn seen to impact on such structures. A particularly clear example of this is the radical statement of Bourdieu: 'The power of words is nothing other than the delegated power of the spokesperson' (Bourdieu, Chamboredon, and Passeron 1991, 107). If language use has no other function then to express power relations, then language use cannot also have an impact on relations of power. The relationship is onedirectional. 1 What still makes interpretive methods important from this perspective is that they help reveal the structural, and thus political, aspects of human interactions. Interpretive methods applied this way do not, however, help us to understand social change, especially not change initiated by self-reflective agents (Bohman 1997). The second option is, conversely, to see social meaning and meaning-making processes as potential generators of structural change. This option gives a greater role to human agency, by downplaying structural conditions. Radical ideas and innovative political narratives may, for example, if forceful, gain sufficient support and following to change the ways societies work. On the downside, such analysis often appears unrealistic and naïve without an account of the interests involved, and at least to be of little usefulness in the more common situation in which power relations and structural conditions still rule the day (see also Rodgers 2012).
Comparing these two options shows that it matters not only if interpretive methods are combined with structural analysis, but also how, more precisely, they are combined. The first of the two options appears unnecessarily deterministic, because subjective meanings are seen as simply the results of social structures. The second appears naïve and counterproductive, because it suggests that ideas and experiences impact on social relations and structures, but does not clarify how they are conditioned by them. Relational analysis offers a way to avoid both problems by recognizing that there is a two-way relation between structures and social interaction. How people experience and make sense of various acts and interactions, in any form and at any site, is central to a relational analysis of interactions in a deliberative system. Likewise, without understanding the structural and institutional conditions that shape such interaction, little can be known about how a site of deliberation is connected to the system and to other sites. Part of what connects sites with each other and with the system lies in the interaction itself, which may consciously or unconsciously reflect various issues and problems of a society, and cause reflection, disruption or reinforcement. Understanding such deliberative impacts and consequences, and thereby understanding the function of events, acts and interactions for the system, requires that we connect structure and social interaction as a dynamic two-way relation.
For deliberative theory, this understanding of meaning as related to institutional conditions helps answering some important questions about a system's capacity for uptake and inclusion. In Ercan, Hendriks, and Boswell's (2017) argument for interpretive approaches, this important aspect of empirical research is toned down, and interpretivism is explicitly distinguished from approaches that seek generalizable insights about deliberative processes. It is said that 'interpretive political research emphasizes contextuality, rather than generalizability,' and that it seeks to 'understand, or make sense of, a phenomenon, in its local, historical and social context' in contrast to 'modernist approaches' that 'aim to develop predictive causal laws ' (198). Interpretive research does not aim for explanations, these authors argue, except in the sense of explaining a phenomenon 'in terms of actors" understandings of their own contexts' (Schwartz-Shea and Yanow, quoted in Ercan, Hendriks, and Boswell 2017, 198). Structural analysis makes it possible to explain, by connecting empirical findings to theoretical and general relations. By contrast, according to the programmatic statements of interpretivism, interpretive research supposedly 'minimizes the role played by the theoretical literature that informs the research question' in order to be faithful instead to the concepts that 'emerge from the field' (Schwartz-Shea and Yanow 2013, 38-39). The cost of embracing this dualistic view is the rejection, not only of 'positivism,' but also the ambition to 'construct and discover' objective conditions of meaningful deliberation, and the critique of exclusionary social structures. 2 There are, however, several examples of research that shows that in practice, one does not need to choose between structural and interpretive analysis. Instead these two perspectives can be fruitfully combined.
Interpretive research that is also relational A relational analysis combines interpretivism's focus on social interaction and meaning with structural analysis that seeks to explain and critically reflect on such interactions and meanings by exploring institutions and structural conditions. This is what many important contributions to research on deliberation have done in the past, including the research discussed by Ercan, Hendriks, and Boswell (2017) as examples of interpretive research. For example, Ercan's (2015) analysis of discourses of 'honor killing' explains culture-based framings in the German debate by taking 'into consideration the institutional and discursive context within which "honor killings" were politicized and debated' (670). Ercan's approach is interpretive in that she explores the ways actors in the deliberative system construct 'honor killings' differently, but she also shows how the case provides lessons for public deliberation more generally; the framings of the subject were conditioned by social structures and political practices, which were reproduced in turn by the discourse of honor killings.
Moreover, Hendriks' (2012) study of four cases of interest-group deliberation, which has also been described as interpretive, makes an important theoretical contribution that goes beyond interpretive analysis. Hendricks argues that strategic action does not undermine, but may in certain circumstances support, deliberative processes. Interests facilitate deliberative interactions, but where they work well they also affect the understanding of such interests and thereby how incentives may influence further deliberation. This, too, is a good example of relational research, because meaning-making processes are shown to be strongly connected to objective, structural conditions. A third example is Boswell's research on 'toxic narratives' (2014,2015). While Ercan et al. are right that this research is an important example of interpretive research, Boswell, too, has contributed to deliberative theory by developing generalizable knowledge about which acts and sites should be seen as important parts of a deliberative system, and this required him to step over to relational analysis. He asks: should wellinformed opinions count as legitimate parts of public deliberation, but not layperson's views and 'toxic narratives' which aggressively oppose the views of experts, misrepresent evidence and vilify vulnerable groups? Boswell's findings illustrate a relational analysis by showing that structures of inclusion and exclusion affect practices of deliberation, and that such practices, in turn, affect the conditions for further deliberation: Rather than necessarily violating the ethical function of the deliberative system, the UK experience shows that inviting adherents to this toxic narrative into elite deliberations can have a transformative effect, encouraging them to moderate the tone and content of their claims to meet the dignified norms associated with such sites. And rather than shutting down debate or leading to dysfunction, the active inclusion of proponents of this narrative into such sites can actually work to reinforce the legitimacy (both real and perceived) of political decisions and the processes that lead to them (2015,326) As these findings illustrate, interpretive research can make important contributions by reconstructing mechanisms that explain unexpected outcomes. But this is precisely what interpretive research, according to programmatic statements, should avoid doing in order to emphasize 'contextuality, rather than generalizability' (Ercan, Hendriks, and Boswell 2017, 198) and 'minimize' the role of 'theoretical literature' (Schwartz-Shea and Yanow 2013, 38-39). Fortunately for deliberative theory, in practice it is what interpretive researchers have done, and successfully so.
Finally, the studies included in Fung and Wright (2003) edited volume on 'empowered participatory governance' explicitly aim to specify similarities in deliberative institutions that have successfully empowered participants and made meaningful interactions possible. The primary aim is to develop a theoretical model for real-world deliberation (called 'empowered participatory governance'). While the contributions to that volume might be considered interpretive in the way they make extensive use of participants' own understanding of deliberative processes, I argue that they are relational in so far as they combine an interpretive focus with an ambition to work out general institutional and structural requisites for empowered participation, and claim that these practices of deliberation have important consequences for the surrounding societies. Fung's contribution, in particular, has these components, although it is not clear to what extent the practices that result from the specified institutional conditions have actual consequences for social structures. His study focuses on citizen participation in school councils and community policing initiatives in Chicago. Describing the practices of deliberation in these cases as characterized by 'accountable autonomy,' Fung argues that institutional conditionssuch as external reviewfacilitated interactions where participants were held accountable to rules of constructive problemsolving, while at the same time being autonomous with regard to outcomes. According to Fung, hence, structural conditions provided the right incentives for citizen deliberation. What is not as clear from the analysis, however, is whether the resultant deliberation had any impact on the structural conditions, an aspect relational analysis would address. Fung claims that his cases are examples of how political reform can take inspiration from 'the traditions of civic engagement and participatory democracy rather than public management techniques or competitive markets,' thereby promoting, besides substantive objectives, the values associated with such forms of engagement and participation (Fung 2009, 9). For research more deeply informed by a relational perspective, a central question would be how these values were promoted through these initiatives and with what consequence. What relational analyses share is a decentered understanding of deliberation, situated through structural analysis, which aims also to trace political and social consequences of the interactions.

Three steps of relational analysis of deliberation
Relational analysis helps address central questions for assessing deliberative systems, including which sites are important parts of a system, how the different sites are connected and what consequences they may have for the system. These are important questions because they should affect our understanding of the system as well as how its capacity for uptake and inclusion can be improved. To make this approach more concrete, I will outline three steps of a relational analysis that help address these questions. In the subsequent, final part of this paper, I will illustrate these three steps by describing a case study of citizen deliberation on urban riots. Here, I will describe them in more general terms.

Decentering deliberative systems
Decentering refers to 'the separation from one another of the objective, subjective and intersubjective worlds qua referential domains of our claims' (Benhabib 2005, 761-762). It means, that intersubjective experiences do not always have to be validated by objective facts first to be regarded as important, and it means that subjective experiences are important even if they are the experiences of only a small group or one person. Local narratives among those who live or work in areas affected by specific problems and issues might differ significantly from media perspective, in which expert opinions and standard journalistic narrative may have larger impact on the discourse. Regardless of whether they are objectively true, such local narratives are politically and socially important. Decentering means that we should not reduce the variety of experiences to one narrative, or our understanding of them to a measure of correspondence with theoretical concepts, but instead try to understand and contextualize the different perspectives on their own terms.
Decentering is an important idea in interpretive research (Yanow 2003, 11). A relational understanding of decentering would add, however, that it is important, not only to interpret local and grassroots discourse, but also to connect that analysis to a structural account of patterns of exclusion from dominant discourse (see also Mendonça 2016). Fraser (1990) argues, along these lines, that deliberative theory should focus on articulations of societal problems that take place in 'subaltern counterpublics,' which reveal structurally dependent practices of exclusion and marginalization. Stevenson and Dryzek (2012) offer another example of such relational reflection in their study of global deliberative governance of climate change, which is shaped discursively through various sites and discourses. Showing the differences and connections between discourses in the deliberative system may, they reason, provoke critical reflection about specific views on climate change, but also on the conditions of pluralistic deliberation itself.

Connect sites of deliberation to structural conditions
The second step is to connect the sites, acts and interactions to structural conditions, and the results of interpretive research to central questions of equality, epistemic uptake and social inclusion. Marginalization, from a deliberative perspective, is a problem of unequal ability to initiate deliberation and affect the direction of debates, due to different positions in the public sphere. Interpretive studies can benefit from drawing on various structural theories of uptake and marginalization as analytical tools to understand the structural conditions for participation. For example, Habermas (1996) locates the obstacles to uptake in public deliberation in the 'systemic distortions' that come from economic, political and other interests that are averse to mutual understanding. When such interests are allowed to dominate, persons become marginalized if they, for example, lack the skills to formulate their concerns in the narrow terms dictated by the market, by political competition or by other spheres dominated by instrumental reasoning.
Another example is offered by Bourdieu's theory of social fields and symbolic power. Several scholars have followed Bourdieu in thinking of exclusion as multilayered/ exclusion as multilayered (see Hayward 2004;Olson 2011;Holdo 2015Holdo , 2016. On one hand, contributing to public deliberation requires some kind of socially recognized competence and authority to speak. On the other hand, the norms that make some personal features recognizable as signs of competence and authority are, to a significant extent, also products of power relations; they tend to be biased in favor of dominant groups' ways of deliberating (Holdo 2014). For example, Calhoun shows that proletarian leaders were excluded from the bourgeois public sphere in 19 th century United Kingdom by referring not to their political views but what was regarded a politically neutral standard of competenceeconomic independence. Working class intellectuals were judged incompetent by claiming that they were unable to rise above class-based interests, unlike the bourgeoisie who, thanks to their material status, could supposedly consider the nation's best interest without the interference of business interests. As suggested by this research, a system's deliberative capacity depends largely on the terms of inclusion and how such terms may change as the result of discursive engagement. 3

Explore deliberative impacts
Engagement in sites of deliberation may impact on a deliberative system in various ways. For example, participants may consciously bring with them a perspective they wish to express, or an issues they wish to address (see Baiocchi 2003). They may provoke deep critical reflection by disrupting routine procedures and exchanges (Piven and Cloward 1979). They may also, of course, not have much other function in the system then to reproduce established ideological preconceptions or reinforce status differences (Hayward 2004). Or, under more favorable conditions, they may start new conversations by bringing together people who share a need to talk about something. Even 'non-deliberative acts,' such as protests, riots or performances, may result in new dialogues, among citizens and between citizens and political representatives. They may raise important issues, or as symptoms of larger problems that need to be addressed. They may, alternatively, themselves be seen as urgent problems to solve. The significance of this last step in relational analysis, is that it means to explore how productive and creative interactions can effect social change and thereby affect the conditions for further deliberation, in which uptake and inclusion will be more effective.

An illustration: deliberation after riots
By way of illustrating the relational approach to studying various interactions that form part of a deliberative system, I will discuss a case study of the Stockholm riots of 2013.. Parts of the results of this research is presented in Holdo and Bengtsson (2017). The study approached the riots as a 'non-deliberative' act, that nevertheless may have very important consequences for a deliberative system (Owen and Smith 2015;Smith 2016).
Riots raise several important questions not only for public deliberation, but also about practices of public deliberation. Which narratives dominate public debates and influence political decision making? Which interests shape the way influential stories about riots are told? Which stories are marginalized? How do local stories from near the events differ from the official story told by policy makers? And what do the answers to these questions tell us about a society's capacity to include and give recognition to the views of all citizens on equal terms? As Horowitz puts it, 'the riot reverberates throughout the political system long after the debris has cleared ' (2001, 11). Riots may lead to 'public introspection, to a determination to see what went wrong and to set it right' (ibid); however, they may also further erode social and political trust, create a wish for revenge and so lead to further violence. What happened in the eyes of citizenswhether, for example, the stories center on 'a mob attack,' 'a clash,' 'a protest,' 'criminal acts' or 'excessive police brutality'will affect the credibility of state representatives and the legitimacy of their responses, as well as people's views of the media channels that portrayed the events and people's relations to other groups whose actions or reactions were somehow important in the chain of events. Riots provide crucial tests of a system's capacity for uptake and inclusion. In the best case, riots lead to critical reflection on problems that have long been neglected. In the worst case, they are fitted and used to reproduce pre-existing narratives and do not lead to any critical reflection about those narratives.
The case study of the 2013 Stockholm riots explored experiences of deliberation through interviews with locals that were affected by them, in which they were asked to give their account of the events as well as the subsequent debates about the nature of the riots and their causes. The riots had started in the relatively deprived suburb Husby, following an incident where the police had fired deadly shots at a senior resident. Over several nights, youths from Husby as well as other parts of Stockholm were involved in confrontations with the police. Vehicles were burnt and public property, including schools, were damaged. In the debates that followed, some framed the riots as a consequence of increased presence of criminal networks, while others claimed that they were symptoms of systemic discrimination, high rates of unemployment, cuts in welfare services, the symbolic distance between the city and suburbs, racism within the police force, lack of integration policy and residential segregation (Adman 2013;de los Reyes et al. 2017). The locals interviewed in the case study often claimed that the riots had to be understood in light of the police killing of the senior resident, but also in the context of marginalization and a widespread sense among youths of constantly being treated by the police as potential suspects rather than citizens to be protected. Peaceful demonstrations had called for a thorough investigation, but neither the police nor political leaders responded to these demands. The riots thus catalyzed debates about several long-ignored problems, including growing economic inequality, urban segregation, stigmatization, discursive marginalization, racism, police-citizen relations and the power of criminal networks in marginalized parts of Stockholm. These apparently nondeliberative acts thereby came to play a significant role in the Swedish deliberative system.
Step 1: decentering citizen deliberation on riots The meanings of riots may vary significantly between different sites of deliberation (Snow, Vliegenthart, and Corrigall-Brown 2007). In this case, the two most dominant understandings represented in media reports and debates, were (a) that the riots were protests against injustices or (b) that they were organized and carried out by criminal gangs. These explanations appeared superficial and weak in the eyes of local civil society activists and persons who worked with social problems in the affected areas. While social injustice and criminal networks were common themes in local narratives, these narratives were far more nuanced and complex. They typically drew together several factorse.g., housing conditions, structural discrimination and the tendency among young boys to look up to local leaders that were involved in criminal activities and who would speak about the police as the enemyto create a contextual understanding of the events that triggered the violence. These stories identified structural problems that were difficult to affect, as well as proposals for how to improve the situation significantly. While most of the local stories could be seen as mutually complementary, rather than competing, they stressed different aspects of urban marginalization. Decentering allowed this study to reconstruct these local narratives, which offered more complex interpretations than did the popular understandings in political, media or academic debates.
Step 2: connect local deliberation on riots to structural conditions The case study on the Stockholm riots analyzed local experiences in order to detect mechanisms of exclusion and marginalization in public deliberation. It asked: why were these perspectives on the riots largely absent in the narratives constructed in the media and political debates? Concretely, interviewees were asked why, in their opinion, the media, politicians and pundits, did not accurately describe the events and their causes. They identified several important mechanisms, two of which may serve to illustrate the way local deliberation can be connected to structural conditions. First, it seemed, they would say, that the story, as told by the media, was already written before the events actually took place. The riots were fitted into an already established plot, with villains, heroes, victims and even lessons to learn. Views that did not fit in, or that did not reinforce this story, were excluded or misrepresented to reduce complexity. Second, there was a tendency that a few individuals were made, or made themselves, representatives of the entire Husby population. This was made possible, interviewees said, by the media's oversimplified idea of the suburb as one homogeneous group of working class people with immigrant backgrounds, with one single perspective. In reality, there were strong disagreements between people in Husby that partly reflected local power structures and struggles over leadership and recognition. In particular, 'the local perspective,' as it was constructed in the media, tended to be a male, leftist, youth perspective, according to which the riots were a protest against injustices. This narrative fitted the two dominant narratives of the riot equally well, it seemed to confirm both that structural conditions caused (or at least motivated) the riots and the equally simple counter-narrative that the riots were organized by extremist and violent-prone youths.
Step 3: identifying deliberative impacts The case study of the Stockholm riots found that, for several actors, the riots resulted in new opportunities to engage in public deliberation. First, the riots helped demonstrate the crucial work done by civil society and community leaders in marginalized suburbs. As emphasized by people interviewed in the study the riots were brought to a stop, not by the police, but by local residents, who came out to talk to rioters and encourage them to express their frustration peacefully. For many individuals, this became an opportunity to get involved in local civil society groups to engage with the problems of the suburb. Public meetings were held to build stronger and larger networks of civic associations and interest groups. New actors appeared to contribute to constructive initiatives. A young entrepreneur, who had watched several cars owned by his business being burned down during the riots, decided to start a youth center in Husby. Other individuals involved in the local civil society argued, however, that the riot was a missed opportunity, where politicians and the media could have done more to address the living conditions and work with people during time with unusual levels of social activism.
The experience of riots had important deliberative impacts. As locals engaged in deliberation with decision-makers and representatives of local authorities to address social problems related to the riots, which were centrally concerned with the relations between local youths and public institutions, they based their sense of legitimacy in part on the consensus view that they had a crucial role to play in improving relations between the local communities and public institutions, including the police. This meant that civil society groups were enabled to play critical roles in deliberation with decision-makers and engage with dominant perspectives through the legitimacy gained through social activism.
The relational approach thus meant decentering deliberation on riots, identifying structural conditions that influenced uptake and inclusion and, finally, identifying deliberative impacts. In this case, the deliberative system failed to connect local perspectives to dominant discourse partly because media narratives were set by remote actors who had preconceived ideas of what story about riots would fit their interests, and because of their simplistic notion of including 'a local perspective,' which served to reinforce local power relations and marginalize alternative perspectives. But the riots also meant new possibilities for change through the engagement of social activists, who were recognized as legitimate partners by local authorities, based on their competent involvement in stopping the riots and creating dialogues between locals and public institutions. By combining interpretive methods and structural analysis, this research could contribute with bottom-up perspectives, generalizable knowledge and draw out political implications and lessons for institutional reform.

Conclusion
This paper has developed a relational approach to research on deliberative systems, which combines interpretive methods with structural analysis. The point of this approach is to identify sites of communication that serve important functions in a system, connect different sites and assess their impact. I outlined three steps of a relational analysis: decentering citizen deliberation, connecting sites of deliberation to structural conditions and exploring deliberative impacts. These steps were illustrated by research on citizen deliberation on urban riots.
I have also sought to show that there are several good examples of research that applies this relational approach to examine deliberative systems. Scholars who have conducted this research have, however, toned down the structural component and overemphasized the interpretive aspects. The important contributions of this research become clearer when the relations between specific interactions and structural conditions are highlighted. For research on deliberative systems, crucial questions demand more than understanding subjective meanings of deliberative interactions. They concern both understanding and explanation. Relational analysis should therefore play a crucial role in analyses of deliberative systems and in eliminating obstacles and distortions to inclusive deliberation aimed, in part, to inform sound decisions that make democratic governance more rational and sustainable.
Notes 1. As Emirbayer (2010, 411) shows, however, Bourdieu's analysis of social phenomena is in practice distinctly relational in its understanding of the interaction between structure and social practice. 2. Important contributions to, and predecessors of, interpretive research would find the rejection of general theoretical statements, and theory-informed sociological reflection, problematic, including Foucault (1977) and Weber (2002). Both sought to unveil social institutions that provide objective conditions for certain kinds of knowledge, ideas and perspectives. 3. Yet another potentially fruitful perspective on marginalization is provided by the term 'discursive opportunity structures ' (McCammon et al. 2007),) and 'discursive fields' (Snow 2004). These concepts, which have recently been used in social movement research, stress that public discourse is not only a target of activism, but also shapes the opportunities for actors to participate and to make certain claims. Public discourse is 'patterned and patterning,' 'enabling as well as constraining,' (Polletta 1999, 67). Whether the goal is a particular political decision or discursive change, actors' opportunities are structured by existing linguistic practices. Such structures may be stable or volatile, institutionalized or not, broad or narrow, and they fit some actors' objectives better than others (see McCammon et al. 2007). Discursive opportunity structures help explain the success or failure of certain actors and certain ways of framing claims in the public sphere.

Funding
This work was supported by Vetenskapsrådet .

Notes on contributor
Markus Holdo is a researcher at the Department of Government, Uppsala University, and currently a democracy fellow at the Harvard Kennedy School.