Ideals of Inclusion in Deliberation

Building on prior thinking about political representation in democratic deliberation, we argue for four ideals of inclusion, each of which is most appropriate to a different situation. These principles of inclusion depend not only on the goals of a deliberation, but also on its level of empowerment in the political system, and its openness to all who want to participate. Holistic and open deliberations can most legitimately incorporate and decide for the people as a whole if they are open to all who want to participate and affirmatively recruit perspectives that would be underrepresented otherwise. Chicago Community Policing beat meetings offer an example. Holistic and restricted forums (such as the latter stages of some participatory budgeting processes) should recruit stratified random samples of the demos, but must also ensure that problems of tokenism are overcome by including a critical mass of the least powerful perspectives, so that their views can be aired and heard more fully and effectively. Forums that aim to improve relations between social sectors and peoples should provide open access for all who are affected by the issues (relational and open), if possible, or recruit a stratified random sample of all affected, when necessary (relational and restricted). In either case, proportional representation of the least advantaged perspectives is necessary. However, when deliberation focuses on relations between a disempowered group and the rest of society, or between unequal peoples, it is often most legitimate to over-sample the least powerful and even to create opportunities for the disempowered to deliberate among themselves so that their perspectives can be adequately represented in small and large group discussions. We illustrate this discussion with examples of atypical Deliberative Polls on Australia’s reconciliation with its indigenous community and the Roma ethnic minority in Europe.

Tensions between equality and equity occur at every stage of public deliberation in civic forums, but perhaps nowhere more than with respect to the question of inclusion. 1Given that deliberative theory is premised on the idea of free and equal citizens exchanging reasons and making decisions together, an abiding concern from both critics and champions of deliberative approaches has centered around whether background inequalities harm disadvantaged groups at various points in the deliberative process (Young, 2000).Such harm may occur prior to any reasons being exchanged at all when inequalities shape who is able to show up to deliberate in the first place.As Gutmann and Thompson put it, "When power is distributed unequally and when money substantially affects who has access to the deliberative forum, the results of deliberation in practice are likely to reflect these inequalities, and therefore lead, in many cases, to unjust outcomes" (2004, p. 48).
A common response to such concerns has been to focus on the representativeness of the deliberating group, making sure that forums are a microcosm of some larger population, whether international, national, or local.One strategy for maximizing inclusion is to open the forum to all who want to participate, while making special efforts to recruit a critical mass of people who would likely be under-represented otherwise (Leighninger, 2012).A second approach, which some deliberative theorists prefer, involves using random sampling to create a deliberating body that looks like the larger group being sampled in as many ways as possible, while giving each member of a population an equal probability of being invited to participate (Barber, 1984;Carson & Martin, 1999;Fishkin, 2009;Gastil, 2008).
Either approach prompts a critical question: who or what exactly needs to be represented inclusively?Many theorists of deliberation have argued that all who are affected by a decision should be represented in discussions (e.g., Cohen, 1989;Dryzek, 2000;Habermas, 1996), while recognizing that in an increasingly interconnected world, it is difficult to draw boundaries around those who are and are not touched in some way by an issue or a decision about it (Fung, 2013;Goodin, 2008).Thus, a deliberation's legitimacy depends, first, on the justifications for defining who is affected by the issues on the agenda (Karpowitz & Raphael, 2014).Assuming that one has defined those boundaries appropriately, it makes sense to think of the population of all affected by a decision as a collection of perspective bearers most relevant to the issue under deliberation.
What are perspectives and why should we focus on representing them inclusively?In deliberation, participants should be open to reconsidering their beliefs, values, interests, and policy preferences, none of which can be assumed to be simple expressions of their ascriptive characteristics (such as ethnicity, sex, income, or sexual orientation).A perspective involves a structural location in society, which can, of course, be closely connected to social identities of various kinds, but in emphasizing perspectives, we seek to avoid essentializing those identities.Perspectives exist prior to deliberation and enrich discussion as they endure during deliberation.Iris Marion Young (2000) defines perspectives as the accumulated "experience, history, and social knowledge" derived from individuals' locations in social groups (p.136).However, as Young makes clear, perspectives do not determine the content of any individual's beliefs, interests, or opinions.One's perspective consists, instead, "in a set of questions, kinds of experience, and assumptions with which reasoning begins, rather than the conclusion drawn" (p.137).In this sense, African-Americans can be said to share a perspective on public life that stems from their common experience of being perceived as black in America.Public forums about racism, income, policing, and many other issues would obviously want to be inclusive and representative of African-Americans without assuming that all black participants will agree on a set of shared interests or values, much less public policies.Similarly, a forum on free speech in schools should aim to include and represent students, who share a similar structural location with respect to the issue, even though they will not necessarily agree with each other about the specific rules, boundaries, and policies that might be proposed in a given school (for further discussion of perspectives, see Karpowitz & Raphael, 2014).
Perspectives are critical to the question of inclusion because a given perspective cannot be easily adopted by someone whose life experiences have occurred in a different social location than that occupied by the holder of that perspective or who has not shared the same set of experiences, history, and social knowledge.Mansbridge argues, for example, that the "vicarious portrayal of the experience of others by those who have not themselves had those experiences is often not enough to promote effective deliberation" (1999, p. 635).Of course, empathy is an important potential outcome of deliberative exchange, but such empathy is less likely if those with a given set of life experiences and perspectives are not present in the discussion.For example, a deliberative forum about contemporary immigration policy would likely be incomplete if it did not include the perspectives of undocumented immigrants brought to the United States from countries like Mexico or children born in the United States to parents who are undocumented.Asking that those immigrant perspectives be fully captured by, say, individuals who immigrated legally from Europe as adults is likely asking for an imaginative and empathetic leap that will be too much for even the bestintentioned deliberator.An effective deliberative system should ensure that all relevant perspectiveholders are heard, and our specific focus is on how the least powerful perspectivebearers can be included in civic forums. 2 As Mansbridge (1999) has argued in relation to legislatures, a critical mass of more disadvantaged perspectives may be helpful for several distinct reasons. 3First, when disempowered perspectiveholders are present, their voices are more likely to be heard, and the stock of arguments, experiences, reasons, and evidence from disempowered perspectives is likely to be larger.Second, and relatedly, a critical mass may bolster the courage of the disempowered to offer minority viewpoints, and hearing those viewpoints from more than one deliberator may push othersboth members of the disempowered group and those who are comparatively more empoweredto take those ideas seriously.Third, in many forums the giving and receiving of reasons occurs not only in large plenary sessions, but also (and crucially) in small break-out sessions and discussion groups, both formal and informal.A critical mass thus increases the likelihood that disempowered perspectives are heard throughout the forum.Fourth, when a greater diversity of disempowered viewpoints is present, no one deliberator is forced to be a token who represents the whole of her social group, a fact that works against an oversimplified, stereotyped, or essentialized view of who the disadvantaged are and what they want.
Avoiding the problems of tokenism is a key advantage of including a critical mass of underprivileged or disempowered perspectives in civic forums.In a deliberative setting, being the only member of a disadvantaged group can be a great psychological and discursive burden (Fiske & Taylor, 1991;Kanter, 1977;Niemann, 2003;Saenz, 1994).Deliberation is a challenging form of communication under the best of circumstances, and all the more so when laboring under the extra weight that comes with feeling hyper-visible and constantly scrutinized, uncomfortably different from others, expected to represent the views of a much larger group, and the subject of others' generalizations about one's group.Tokens often feel they must devote additional attention and energy to impression management, adjusting their speech and behavior to both stand as positive exemplars in the eyes of their in-group while simultaneously avoiding negative stereotypes or other adverse judgments from the out-group.In reaction to these stresses, tokens who represent disadvantaged populations may pull back from full engagement with others.For example, research shows that women who are gender tokens in discussion groups often speak less, are less likely to receive positive feedback from men, and are seen by others as less authoritative discussants, while male gender tokens are not hampered in these ways (Karpowitz & Mendelberg 2014).
Overall, then, a critical mass of disadvantaged voices is likely to lead to a more complete exchange of deliberative reasons.
The diverse views of the disadvantaged can be heard more fully; more privileged members of the deliberating group are more likely to take those views seriously; and a civic forum is more likely to communicate to both participants and non-participants that all voices are valued and that the community as a whole has been well-represented.

Four Ideals of Inclusion
If deliberative ideals are furthered when the voices of the disempowered are included, the standards and specific forms of such inclusion are likely to vary with the goals of each forum, the kinds of issues it considers, how it relates to formal decision-making power and institutions, and its level of access to potential deliberators. 4Our focus here is on two elements or dimensions of civic forums: the goals of the forum and its approach to access.The interactions between these two criteria produce four potential ideal standards of inclusion.
We begin by considering the goals of civic forums.Some aspire to a holistic vision in which they aim to represent the collective opinion of, or make decisions for, an entire people (demos).Often, these sorts of forums aim to explore collective goods, the management of common resources, or the design of democratic institutions and decision-making procedures.For example, holistic forums have taken up issues like collective budgeting, infrastructure plans, visions of how future development might proceed in a location, the management of public lands, public safety and policing, and even constitutional issues like a nation or region's electoral system.Sometimes these sorts of forums are empowered to enact policy directly, such as Participatory Budgeting or community policing.They might even be built into traditional institutions, as in the case of the New England Town Meetingone of the oldest and most venerable civic forums in existence today.Other forums serve an advisory role to existing decision-making institutions.Regardless of its specific level of empowerment or topic, though, these sorts of holistic forums often take up issues that affect nearly all parts of the public deeply and regularly.
Other forums focus on relational mattersspecifically, how one sector of the public relates to another or to the public as a whole, or even relationships across borders and between different publics.The goal, then, is to make decisions, or advise decision makers, about issues that might affect some portions of the public more frequently or more deeply than others.These forums often tackle social problems, minority-majority relations, relations between communities, and transnational issues.The large and growing number of inter-group dialogue programs, such as the community dialogue groups about race studied by Katherine Cramer Walsh (2007), also exemplify this approach.Other examples might include issues like alcohol abuse or domestic violence, tension (and potential reconciliation) between religious or ethnic groups, or the condition or status of certain minority groups.Often, these sorts of forums serve an advisory role and are not designed to make decisions directly on behalf of the full community.
Beyond the goals of a forum, a second dimension is the forum's level of openness to all who want to participate.Some forums invite everyone who wants to participate, while others restrict participation to smaller numbers who represent broader groups of citizens.This distinction between open access and more limited access has implications for strategies of recruitment and, specifically, methods for incorporating disempowered groups.
Table 1 summarizes ideals of inclusion for four different types of forums, based on the forum's goals and access.For each type of forum, we identify a standard for recruiting an inclusive sample and give a real-world example of that standard.

Holistic Open Forums
Quadrant 1 of In open forums of this sort, every citizen has a formally equal opportunity to participate, but self-selection means that not all are likely to actually do so.Given background inequalities in resources and civic skills, forums that rely on self-selection are likely to over-represent people who have more time, money, and education as well as greater stores of political efficacy, information, and motivation (Verba, Schlozman, & Brady, 1995).Those individuals with greater resources and more politically engaged social networks are also more likely to be asked to participate (Rosenstone & Hansen, 1993).In practice, this means that older people, males, and members of more privileged racial and ethnic groups are likely to be disproportionately represented (Lee, 2011;Ryfe & Stalsburg, 2012;Steiner, 2012).
With these shortcomings of self-selected samples in mind, a forum that aims to represent all voices may thus affirmatively recruit or otherwise seek to mobilize the least powerful.Among other possibilities, such efforts might include contacting networks of associations that can reach under-represented groups or simply advertising opportunities to affect policy on issues of special importance to the least privileged (see, for example, Lee, 2011).If such opportunities are genuine and if recruitment is successful, then the resulting forum can be one in which all relevant perspective-bearers show up and have not just a formal but a substantively equal opportunity for inclusion.
Full inclusion in these types of forums is often most feasible in neighborhoods or small towns-localities that are small enough to make including everyone a realistic goal. 5An example is the Chicago Community Policing meetings that brought together citizens and public safety officials to make collaborative decisions about improving neighborhood crime prevention (Fung, 2004;Skogan & Hartnett, 1999).These forums were empowered to make consequential, onthe-ground decisions about local policing strategies and priorities.Meetings were formally open to all residents of the city's 280 neighborhood beats, which meant that everyone who chose could attend, and the principle of equal access was met.
Because beat meetings were also held in poor neighborhoods, they attracted lowincome residents whose lives were most directly affected by street crime.These meeting participants had, in many cases, been the victims of crime themselves and were thus disempowered on this specific issue.In many cases, they were also disempowered with respect to the broader political system.Because the forums focused on an issue of special importance to low-income neighborhoods struggling with crime, however, the topic of the beat meetings not only represented a powerful argument for including the perspectives of this disempowered group but also functioned as a mobilizing incentive for the least powerful to show up.And, in contrast to many public meetings and other forms of civic engagement, residents of disadvantaged neighborhoods did show up, often at rates higher than those in more affluent areas.By 1998, around 10% of Chicago residents had attended at least one meeting, though a smaller group of regular participants showed up far more often.Thus, the Chicago meetings did not entirely avoid the problem of self-selection and its attendant lack of representativeness, but by holding meetings in all neighborhoods on a topic of special concern to struggling and disadvantaged city residents, these meetings attracted a critical mass of residents who were under-represented in Chicago's elections and other opportunities for political engagement.
The Chicago example also suggests, however, a limit to this approach to inclusion.Equal access for all affected is possible at the neighborhood or smalltown level, but it may be more challenging in forums that take on a wider scope.At the same time, however, the Chicago meetings showed that inequalities in participation that are common in other settings (Verba, Schlozman & Brady 1995) can be offset, at least to some extent, when meetings are held in locations that allow easier access for the disempowered and when the agenda of those meetings addresses issues that disempowered groups see as relevant and meaningful to their lives.

Relational Open Forums
Quadrant II of Table 1 includes forums that are open to all citizens but that address relational issues, such as social problems and interactions between communities or social groups (ethnic or religious groups, for example).As in Quadrant I, these sorts of forums seek to maximize equal access for all, but because they are focused on communication across community or sub-group boundaries, they have a special concern for achieving proportional or even overrepresentation of those most directly and deeply affected by the issue.
An example of this approach to inclusion occurred in 2002-2003 in West Virginia, where a coalition of voluntary associations working to reduce domestic violence held a series of deliberative forums around the state (see Fagotto and Fung, 2006).The goals included increasing public understanding of the issue and identifying the public's preferred strategies for dealing with the problem.Forum organizers developed three major approaches to the problem: increasing community support for victims, stiffening punishment for perpetrators, and altering public attitudes toward domestic violence.The forums were designed to promote discussion of the specific policy steps associated with each approach.
The meetings were open to all, but given how organizers framed the topics, they made special efforts to recruit perspective-holders who were most deeply affected by the issue, including victims, perpetrators, abuse prevention professionals, and law enforcement personnel.These groups all deliberated together in small forums, and by over-sampling people of each perspective, forum organizers increased the likelihood of having multiple opinions within each group rather than relying on only one or two representatives to articulate the views and experiences of the broader group of victims, perpetrators, or others.
These forums were, therefore, committed not only to cross-cutting talk across the different relevant groups, but also to ensuring that enclaves of each perspective were present in large enough numbers that the diversity within each perspective could emerge.In addition, forum organizers followed up these meetings with additional forums designed to discuss the issue of domestic abuse in especially disadvantaged communities, such as people with disabilities, African-Americans, immigrants, and lesbian, gay, or bisexual individuals.In this way, organizers incorporated less powerful perspectives both in the initial forums and again after the forums in successive rounds of additional consultation.
Thus, while this series of forums departed from a strict vision of "representativeness" in the sense that the deliberators at various points were not a random sample of the public at large, they served the democratic and deliberative goal of full discussion among those who were most affected.This approach to inclusion increases the burden on forum organizers to identify and reach out to all affected sub-groups.In some cases, identification may be difficult at the beginning of the process, and additional relevant perspectives may emerge during the discussion, which may mean expanding the process to make sure those perspectives are fully heard, as well.A related challenge is giving sufficient space to all affected sub-groups without excluding the larger, more "representative" community.When the goal of the process is dialogue across sub-group boundaries, not community decision-making, as is often the case with forums that prize relational goals, then departures from strict representativeness may not be cause for concern because the key is communication between relevant perspectives.But when forums lead to collective decisions that are binding on all members of the community, then the connection between affected groups and other parts of the community is critical.In the West Virginia case, because the forums were only one phase of a larger process of developing recommendations for binding collective decision-making, the over-representation of affected groups was not a democratic liability and probably represented a distinct democratic strength.

Sampling in Restricted Forums
Quadrants III and IV of Table 1 represent a different approach to inclusion, one that applies to forums in which it is not possible to open the doors to all and still preserve the possibility of meaningful deliberation.In these cases, the exchange of perspectives may be best facilitated by limiting admission to a select group who represent other citizens.Often, the need to limit access occurs when forums attempt to address large-scale social issues that affect groups much larger than could be accommodated with an open invitation.In such cases, the answer is likely to be some form of representative sampling.But it is only possible to speak meaningfully of a representative sample after defining the population from which to draw representatives and after making decisions about which characteristics ought to be represented.We have argued that these decisions are most legitimately inclusive when organizers think in advance about which perspectives are most relevant and affected by the issue under discussion, and then include such perspectives in appropriate proportions (see also Moscrop & Warren in this issue).
Which proportions are appropriate to serve the goal of inclusive equality?We have noted that many theorists and empirical researchers advocate random selection, which ensures that each member of the population has an equal probability of participating.Proponents of random selection contend that it is likely to yield deliberators who are less invested in the usual political battle lines or less committed to a pre-defined set of outcomes than would occur through selfselection, the electoral process, or network recruiting alone (Hendriks, 2011).
But even the most rigorous approaches to random selection are likely to result in departures from the ideal, for at least two reasons.First, recruiters to most civic forums seldom begin with a list of the population as a whole.Instead, they tend to rely on voter lists (which exclude the unregistered) or random-digit dialing (which may omit voters without telephones or the increasing number of individuals without a land line).Second, even if all are given an equal opportunity to participate through random selection, the decision to take part in civic forums introduces again the challenge of self-selection and its attendant inequalities.
Beyond those challenges, however, random selection may, except in very small locations or very large forums, conflict with representative selection.Put differently, a focus on giving each citizen an equal probability of being chosen does not necessarily translate into an equal opportunity for all relevant perspectives to be represented at the forum, which is, we have argued, the key for effective and inclusive deliberation.Randomly choosing a small number of deliberators from a very large population can result in all sorts of departures from representativeness and, more troublesome still, departures that would pose problems for deliberative equity.Random sampling is especially likely to miss potential deliberators from small minority groups.Consider civic forums like the Consensus Conference or Citizen Jury formats, which rely on a very small number of deliberators-often 25 or fewer-to represent the views of much larger groups, sometimes even as large as a nation-state.Even under the best of circumstances, effectively reflecting the variety of relevant perspectives contained in the larger polity would be a challenge for such a small group.In these cases, formal random selection may even exacerbate the problem.For example, if attempting to select two dozen Americans at random from a larger population of 315 million to participate in a forum about abortion in the United States, it is quite possible that random selection would result in a group composed entirely of men or women, Republicans or Democrats, wealthy or poor.Even if the group were chosen by the most rigorous approach to randomization possible, it would be difficult to argue that a forum on abortion composed entirely of men would sufficiently (let alone equally) represent the perspectives relevant to effective deliberation about the issue.
These problems are not limited to small civic forums, either.Even in much larger forums, such as Deliberative Polls or 21 st Century Town Meetings, where hundreds or even thousands of deliberators are recruited, deliberation does not occur in the large plenary meetings only.Much of the real give-and-take of discussion occurs in small groups, both formal and informal, and a reliance on random sampling alone cannot ensure that a critical mass of relevant perspectives can be present in large enough numbers to influence these smaller discussions.
The remedy for cases like this, in our view, is stratified random sampling, with an eye to ensuring that all relevant perspective-bearers are present-and better yet, present in sufficient numbers to disperse into small-group discussion without being isolated tokens.Such an approach means, of course, that organizers must make crucial decisions about which perspectives are most relevant (and, thus, which perspectives on which to stratify).One objection might be that these sorts of a priori choices are so consequential that they give organizers too much control over the dimensions of deliberative conflict instead of allowing disagreements to arise organically from the deliberators themselves.But such a concern overlooks the fact that organizers always make consequential decisions about the boundaries of relevant publics, the framing of the issues to be discussed, and the composition of those who will be at the meeting.This is especially true of forums where it is impossible to accommodate an entire public.
Even so, random selection may still be a valuable element of equal inclusion.Even if it is true that sponsors and organizers of civic forums are likely to play a key role in defining the relevant population, identifying key perspectives, and choosing the criteria for representing the populations and perspectives, random selection within strata can ensure that organizers do not directly control which particular individuals are chosen.This approach ensures that organizers do not play too intrusive a role, whether consciously or not, in determining who, exactly, represents the larger group, and it guards against organizers imposing their views on who would or would not most authentically represent a social group's interests, or even what those interests might be.Another option for preventing excessive control by organizers might be to allow social groups or strata within a larger population to choose their own representatives through election.

Holistic Restricted Forums
Quadrants III and IV of Table 1 summarize ways of putting principles of inclusion into practice when open access for all is not possible.Holistic forums that must rely on some version of representation (Quadrant III) should aim for a stratified random sample of all perspective-holders affected by the issue, with special concern for ensuring that under-represented populations are present in sufficient numbers to be heard effectively.When random sampling by strata is not possible, an alternative would be asking citizens themselves to choose representatives, but again within strata designed to ensure that all relevant perspective-holders are present.
An example of this approach is the Participatory Budget process first attempted in 1989 in Porto Alegre, Brazil.Though the process has changed to some extent over the years (Baiocchi, 2005;Baiocchi, Heller, & Silva, 2011;Wampler, 2012), the core idea was that citizens attend open meetings in one of the city's 16 districts to review the previous year's budget and to elect fellow citizens to represent their neighborhood in the upcoming year's budget negotiations.The number of representatives for each neighborhood was based on attendance at the neighborhood meetings, a fact that gave citizens, including the least privileged, a strong incentive to show up.Because the city committed itself to a funding formula that allowed for larger public investments in infrastructure in the poorest neighborhoods, the least well-off had an added incentive to take part in the meetings.For all of these reasons, attendance at the neighborhood meetings was robust, sometimes numbering over a thousand residents, including those meetings held in the most disadvantaged places.
The neighborhood meetings tended to be too large for meaningful deliberation, but they were moments of accountability for both city officials and neighborhood representatives.The real deliberation occurred among the 40-60 delegates from the neighborhoods who met multiple times over several months to discuss each city district's priorities.These district councils decided on spending priorities and elected a smaller group of representatives from among them to be part of a citywide budget council tasked with harmonizing the budget priorities of the various districts.City officials joined the meetings at each stage and, in the end, often adopted citizens' priorities in the municipality's final budget.
In essence, the Participatory Budget process was a stratified sample-the district council was stratified by neighborhood, and the city budget council was stratified by district.The process was thus committed to ensuring that choices about the provision of public goods and other infrastructure projects reflected the needs and priorities of the strata (the neighborhoods and districts).Though it did not make use of random sampling, the institutional design ensured that elected citizens represented geographic constituencies proportionally and included a critical mass of groups that had been historically under-represented in city government, such as less-educated people, women, and blacks (see Baiocchi, 2003). 6The process thus delivered on its promises of greater political voice for historically disempowered perspectives and an opportunity for autonomous political action on issues of onthe-ground importance to the lives of those who were least well-off.Because the process began at the neighborhood level, it included opportunities for discussion within strata that had traditionally been disadvantaged, as well as cross-cutting deliberation across the multiple strata that were present in the city.
The challenge of this approach is that the institutions through which representatives are to be elected or selected can exert considerable influence on any subsequent deliberation.The inclusiveness of deliberation and the success of binding decision-making that occurs later in the process depends on institutions that allow all affected perspectives and, especially, traditionally under-representative perspectives to be heard.But the importance of attending to these institutions is not unique to deliberative civic forums; it is present in any approach that relies on representation.The Participatory Budget process shows that representative institutions that are committed to both inclusiveness and deliberation are possible and that they can bring the concerns of traditionally disempowered perspectives to the process (and even allow disadvantaged groups to set the budgetary agenda) in ways that random selection or even citywide voting would have been unlikely to do.

Relational Restricted Forums
Similar approaches to inclusion can be applied to forums with relational goals (Quadrant IV of Table 1), with the primary difference being that in relational forums, proportional or over-representation of perspectives most affected by the issues being discussed is even more important.The need for a critical mass of the least powerful is especially pressing in forums taking up claims that some groups should be treated as distinct peoples within a larger state, as in the case of many forums about the status of indigenous groups, or when the forum is designed to explore how one demos relates to another, as in the case of forums meant to discuss multinational or transnational issues.In such forums, the groups most affected by decisions are often more clearly delineated, and a failure to include representatives of such groups in sufficient numbers to be present in both large plenary sessions and small-group discussions would seem to reject their claims for distinct treatment before the forum even begins.In such forums, these groups may have a legitimate need to be over-represented in relation to their proportion of the larger population.
Two Deliberative Polls that took up claims of politically disadvantaged minority groups illustrate how inclusion of such groups can be achieved in practice.A 2007 Deliberative Poll on the treatment of the minority Roma ethnic group in Bulgaria achieved inclusion using a stratified random sample with proportional representation of relevant groups.The Roma made up about 10% of Bulgaria's total population, and organizers stratified their sample to ensure that a similar proportion would be present at the forum.Because the forum was large, with about 255 participants overall, the Roma were present in sufficient numbers to ensure their participation in all the small-group discussions (Fishkin, 2009).Though concerns may linger about the challenge of tokenism in the small-group setting, both the size of the Roma population and the size of the Deliberative Poll meant that no group discussions occurred in the absence of Roma perspectives.In this case, proportional representation of each stratum (with strata defined by relevant social groups) achieved a high level of inclusion and thus furthered the relational goals of the forum.
A 2001 Deliberative Poll conducted in Australia highlights a second possibility, one where proportional representation was insufficient and over-sampling of the least powerful was needed.The poll took up the issue of reconciliation between indigenous and other groups in Australia, but the indigenous made up only about 2.5% of the population, meaning that proportional sampling likely would have produced only very small numbers of indigenous participants, despite the fact that their interests and concerns were at the core of the forum's purposes.For this reason, oversampling of indigenous populations was required.In addition, organizers went one step further and convened regional discussions among enclaves of indigenous citizens prior to the beginning of the Deliberative Poll.These enclave discussions worked to identify policy strategies for reconciliation, to craft questions to be asked of experts during the Deliberative Poll, and to recruit indigenous people to participate in the poll.
At the national poll, organizers took one additional, very instructive step; they randomly assigned multiple indigenous participants to around half of the small groups.This choice ensured a critical mass of indigenous peoples in those groups, rather than token representation, while the other half of small-group discussions occurred without any indigenous participants.Organizers found that all groups became more supportive of reconciliation policies, but the movement in favor of such policies was significantly larger in groups with indigenous representation.Researchers attributed this difference to "first-hand personal stories of disadvantage from the indigenous Australians, as well as first-hand views of what indigenous Australians want to see done about the disadvantage, both symbolically and practically" (Issues Deliberation Australia et al., 2001, p. 49).This result is strong evidence for the effect of a critical mass of disempowered perspectives in deliberative settings.
Interestingly, the inventor of the Deliberative Poll, James Fishkin, questioned the decision.Fishkin (2009) wondered whether compromising the principle of proportional representation was necessary, given that even participants in small groups without any indigenous peoples became more supportive of reconciliation efforts.For that reason, Fishkin writes, perhaps it would have been sufficient to preserve proportional representation and simply include indigenous perspectives in briefing materials and among experts who participated in plenary sessions.But this position seems to assume that some movement in favor of indigenous preferences is the rough equivalent of the much greater movement seen in groups with indigenous representation, and it also overlooks the potential effect of the presence of a critical mass of indigenous peoples in the large plenary sessions.It remains a matter for empirical testing to see if arguments in briefing materials and the presence of indigenous experts would produce the same effect on their own, when a critical mass of indigenous peoples is not present in the larger deliberating group.

Conclusion
In this article, we have argued that the path to equal inclusion will vary with a forum's goals and ability to accommodate all relevant deliberators.That is, the details of inclusion ought to depend upon the forum's place in the larger deliberative system.Forums that aim to speak for the people as a whole should practice inclusion by opening their doors to all while simultaneously being sensitive to the effects of self-selection, making sure that those who are least likely to participate are affirmatively recruited.When such holistic forums cannot accommodate everyone and therefore must limit participation, a stratified random sample that ensures a critical mass of those who are usually under-represented in public deliberation is the key to ensuring that all voices are heard sufficiently.Forums dedicated to relations between groups, sectors, or peoples can provide equal access for all affected or a stratified random sample of all affected, depending on the forum's ability to accommodate all relevant deliberators.In many cases, proportional representation of the least powerful perspectives will be sufficient.In some conditions, such as when the relevant group is small and the forum's topic directly concerns that group, over-sampling of the group may best serve the cause of full and inclusive deliberation.Such over-sampling can avoid the problem of tokenism (or outright absence from the discussion) and help the voices of the least powerful be heard.
Elsewhere in this issue, we argue that offering opportunities for the disempowered to deliberate among themselves can be a powerful recruiting tool, as well as a legitimate stage of a larger, more diverse public discussions. 7This kind of enclave deliberation among the least powerful can be profitably incorporated into all four of the methods of inclusion we have described here.The forum itself may be an enclave of disempowered perspectives, as in the case of low-income neighborhood residents at Chicago Police Beat meetings who were able to work with public officials to tackle approaches to crime on the ground, in the places where crime had a large effect on the daily lives of the deliberators.In addition, enclaves can be included as one step in a larger process that also includes crosscutting deliberation, as in the case of Porto Allegre's Participatory Budgeting, where neighborhood meetings preceded district councils, which then sent representatives to the citywide budget council.Enclave deliberation can occur prior to the forum in ways that inform it, as in the case of the regional meetings with indigenous peoples prior to the Australian Deliberative Poll.Or enclave deliberation could occur after a cross-cutting forum has occurred, as in the case of West Virginia's meetings on domestic violence with populations uniquely affected by the issue the civic forum had addressed.
The larger point is that ensuring equitable inclusion in practice is often far more complex than merely opening the doors of the forum to all comers or even than inviting a simple random sample of the public at large.Though we are hardly opposed to a role for random sampling or the idea of open access, these important approaches need to be supplemented by careful thought from forum organizers about the goals and purposes of the forum, its ability to accommodate all who might want to show up, who is most affected by the issues under discussion, and the perspectives of those who have traditionally been disempowered with respect to these issues and by the political system as a whole.
Table1focuses on forums with holistic goals and open accessthat is, forums that seek to represent the interests and opinions of all and that can accommodate everyone who wants to attend without making deliberation impossible.The relevant ideal of inclusion in this context involves throwing open the doors of the forum to everyone who wants to attend, while simultaneously making special efforts to recruit those who are least likely to come.