Psychological Mechanisms of Deliberative Transformation: The Role of Group Identity

Democratic deliberation has been shown to lead to shifts in people’s preferences for particular issues. The psychological mechanisms that underpin such shifts are not well understood. Against the backdrop of a deliberative forum we examined participants’ preferences for various types of political systems, how these preferences changed as deliberations proceeded and how the final preferences were associated with different levels of inclusiveness of a social identity. The results showed that at the end of the deliberations people’s preferences moved in the direction of satisfaction with the political system, and that this preference was positively associated with identification with the superordinate identification but negatively associated with the subgroup identification. We discuss the implication of these results for the design of deliberative forums as well as the role of social identity in deliberative democracy. on ballot initiatives and referenda.


Psychological Mechanisms of Deliberative Transformation: The Role of Group Identity
Empirical research has shown that deliberative outcomes have a wide array of effects on those who take part in deliberation, the benefits of which are supposed to produce better deliberative citizens (Pincock, 2012). Deliberation can, for example, reshape participants' policy preferences (Niemeyer, 2011), attitudes (Fishkin, 2018), and cognitive integration (Gastil, Black, & Moscovitz, 2008;Gastil & Dillard, 1999).
However, there is increasing recognition that deliberation is more than the summation of individual effects-there is something going on at the level of the group. The idea of the deliberative group has been subjected to valorisation as well as pathologizing. On the latter, effects such as group polarisation and groupthink have provided cautionary tales of the potential for deliberative pathologies (Stokes, 1998). On the other hand, the mechanisms that correct these outcomes (e.g., Lindell et al., 2017), and benefits of group cognition in deliberation (e.g., Mercier & Landemore, 2012) attract growing appreciation.
The mechanisms that produce these effects during deliberation are characterised, among other things, in terms of facilitation (Mansbridge, Hartz-Karp, Amengual, & Gastil, 2006), prescriptive rules and norms about listening, communication, and interaction that promote a critical evaluation of arguments as well as of one's own position (Dryzek, 2000). In addition, deliberation should lead to shared representations regarding the issue at hand (Dryzek & Niemeyer, 2006. Another dynamic that also warrants attention concerns the dynamics of group identity (e.g., Felicetti, Gastil, Hartz-Karp, & Carson, 2012). Although identity is implicated in recent growth in populist politics (e.g., Marchlewska, Cichocka, Panayiotou, Castellanos, & Batayneh, 2018), this paper draws attention to those processes whereby deliberation produces a qualitatively different identity outcome. Drawing from the social identity perspective (Tajfel & Turner, 1979/2010Turner, Hogg, Oakes, Reicher, & Wetherell, 1987), we will attempt to understanding the role that group identification plays in the deliberative process and its outcomes. To this end we used a deliberative forum and analysed participants' understandings of democracy and how these understandings changed as a result of deliberation and group identification.

Deliberation, Group Identity, and Self-Categorisation
Social identity refers to the parts of the self that are derived from one's psychological membership in social groups. Self-categorisation is the process through which people, by identifying with a social category, come to perceive themselves as having psychological attributes interchangeable with other members of that category. Assuming that deliberative forums exhibit typical group processes, they are likely to induce a shared identity and associated common goals and beliefs among participants (Felicetti et al., 2012;Hartz-Karp, Anderson, Gastil, & Felicetti, 2010). These ideals are consistent with the conception of shared group identity developed in the social identity perspective, according to which ingroup members come to share norms and beliefs that define their group (Turner et al., 1987). As such, if the group is defined by a critical stance and openness to others' standpoints, then these norms become defining of group members themselves.
It is important to note that people's identities can exist at different levels of inclusivity from the personal to the social, with the latter including group identities that can be more or less inclusive (Turner et al., 1987). Social identity theory conceptualizes a hierarchy of inclusiveness that goes from the subordinate to the superordinate level (despite the terminology these levels do not indicate any kind of primacy). A superordinate social identity is one that includes many other social identities. For example, the social category Human (superordinate) includes scientists (less inclusive social identity), which in turn includes social scientists (still less inclusive), which in turn includes a particular person (personal identity, least inclusive).
Self-categorisation at a higher-order level does not mean that personal or subgroup identities switch off. It only means that a particular social identity is salient and therefore has a stronger impact on self-perception and selfdefinition, which then shape behaviour, attitudes and intentions. Indeed, it is possible for the person to simultaneously self-categorise at both the subgroup and at the superordinate level. Research on intergroup relations, for example, has shown that attempts to ameliorate intergroup conflict are better served by promoting a dual identity (e.g., woman and scientist) rather than a single higherorder identity shared across conflictual groups (Dovidio, Gaertner, & Saguy, 2009). This is because higher status groups (e.g., men) tend to project the characteristics of their subgroup into the higher-order group (e.g., scientists) and see themselves as more prototypical of the ingroup than the lower-status subgroups (Wenzel, Mummendey, & Waldzus, 2007). In this vein, research shows that shared higher-order identity can counteract social change by reducing engagement in collective action in favour of disadvantaged groups (Saguy, Tausch, Dovidio, & Pratto, 2009).
Although deliberative theory focuses on shared representations, it also posits that deliberative practice should be able to accommodate difference (Dryzek, 2000). In social identity terms, deliberation should promote a sort of group dynamic wherein participants maintain contextually relevant subgroup identities while merging with a super-ordinate identity. For example, during deliberations, a participant might be aware of her/his identity as a Northern Territorian (NT) in the wider context of also being Australian and what that means. The maintenance of a dual identity during group deliberation is likely to allow deliberators to be critical of the issues being discussed, at the same time as being able to converge with others such that goals that benefit the common good, rather than narrow self-interest, can be achieved. In keeping with the example above, the NT participant would be able to see issues that are important for Australians as a group and, at the same time, provide input using a NT lenses to impart a critical subgroup perspective. Such conditions are more likely to promote a dialogue oriented toward social change. In this investigation we will test these ideas and examine the content of understandings of democracy, whether they change as a result of deliberating about it and how such changes are related to social identification at different levels of inclusiveness.

Research Overview
These ideas are tested in the context of the Australia Citizen's Parliament (ACP), a large-scale deliberative undertaking that brought together 152 Australians from each federal electorate in Australia (Carson, Gastil, Hartz-Karp, & Lubensky, 2013). 1 Using deliberative minipublics of this type and the (ideally) open-ended nature of potential outcomes limits their use as field experiments. One of such limitation is that it prevents the researcher from hypothesising in advance the kind and direction of change, including the level of social identification. Thus, this investigation is more exploratory rather than probative.
Nevertheless, based on the design of the ACP, assumptions and empirical findings from deliberative forums, and experimental work on social identity, we anticipate that, as a result of deliberation, participants change their understandings of the topic of discussion which, in turn, are likely to change beliefs, attitudes and political stance. Because participants in the ACP take part, both as representatives of their electoral district as well as of the nation, we also expected identification, both at the subgroup (electorate) and at the higher-order (Australian) group level to be salient. We also expected group identification at both levels to be associated with understandings of democracy at the end of the deliberations. However, the direction and relationships between these variables are not hypothesised. 1 In order for the sample to be representative of the whole Australia we aimed at having one participant from each of the 150 federal electorates. Initially, 9653 invitations were sent to citizens across electorates. From these 2762 Citizens registered their interest to participate in the ACP. The final 152 (2 extra due to oversampling) participants were drawn from this sample using stratified random sampling. This technique was used in order to obtain a representative sample in terms of gender, age, education and indigenous/non-indigenous citizens. This sampling resulted in 75 men and 77 women, the ages varied between 18 and 90 years. Of these, four participants identified as indigenous. Most participants were between 45 and 54 years old. Education ranged from compulsory education to post-graduation. Most participants had either TAFE (training and further education) or a bachelor's degree.
To track change in understandings of democracy the study was conducted longitudinally covering four stages of data collection (see Figure 1). The participants deliberated on the question, "How can Australia's political system be strengthened to serve us better?" 2 The Citizen parliamentarians' (CPs) task was to discuss this issue and come up with consensually agreed recommendations to the Government. These recommendations were then delivered to the Prime Minister's Cabinet (see Dryzek et al., 2009 appendices 5 and 6 for procedural details). The final meeting of the ACP was held in Old Parliament House in Australia's capital city, Canberra. The four-day meeting was chaired by two prominent public figures, attended by Federal politicians and other key-speakers expert in the field of Australian politics. Deliberations were moderated by facilitators that made sure that deliberative norms were followed. The facilitators themselves were mainly professionally trained and were carefully briefed, in the lead up to the event, in respect to their roles (see Newdemocracy, 2009). Prior to the beginning of deliberations proper the ground rules for discussion were presented. These were: speak openly and honestly; listen carefully to what others have to say; treat everyone with respect; keep comments brief and to the point of the question; stay on task; and if you need to take a break, do so. Previous findings from the ACP suggest that participants followed these general guidelines (Hartz-Karp et al., 2010).

Q Method
We traced the discourses of democracy and analysed them through Q methodology. The Q method is both a qualitative and quantitative form of analysis and is normally deployed to study people's subjectivity (Brown, 1993;Dziopa & Ahern, 2011). It reduces a variety of points of view on a topic of interest into a manageable set of shared views through Q sorts. In this study it was used to extract narratives (i.e., understandings) about Australian democracy, which served as benchmarks for tracking the positions of participants during different stages of the deliberations.
The Q method serves to disclose shared points of view on a topic of interest, such as policy and planning issues (van Eeten, 2001;Ellis, Barry, & Robinson, 2007;Ockwell, 2008), democracy (Jeffares & Skelcher, 2011), or sustainability (Barry & Proops, 1999). Studies aimed at understanding the effect of deliberation often deploy the Q method (Cuppen, 2012;Pelletier, Kraak, McCullum, Uusitalo, & Rich, 1999;Raadgever, Mostert, & van de Giesen, 2012) as it allows for a rigorous evaluation of learning effects of deliberation (Cuppen, 2012). It begins by drawing a sample of statements that represent the issue or phenomenon under study. The participants sort these statements into a predefined grid resembling a normal curve and graded on a scale from -5 (Most disagree) to +5 (Most agree). This technique forces participants to prioritise and highlight the issues that are most important to them. The organisation of the statements in the grid represents personal points of view and permits a comparison with the point of view of others.
Q sorts are then correlated and analysed through inverted factor analysis in order to uncover similar statement organisation. This procedure groups similar Q sorts into clusters, or factors, by computing a weighted average Q sort for each factor. The factors represent ideal type positions around which all other Q sorts "organize." The individual loadings in a factor give an indication of the "distance" between a specific Q sort and the factor. A factor can also be seen as a hypothetical individual, whose beliefs perfectly match the factor in question. Factors are represented in the form of narratives drawn from the underlying statements. Q methodology is based on the idea that, if the factor analysis uncovers patterns in the organisation of statements, then there is a "logic" that accounts for these patterns (see Niemeyer, Ayirtman, & Hartz-Karp, 2013).

Establishing the "Deliberativeness" of the ACP Process
To assess whether participants understood the process as deliberative, we used a direct measure (Black, Burkhalter, Gastil, & Stromer-Galley, 2010) in the form of a questionnaire administered after deliberations had ended. Following Burkhalter, Gastil, and Kelshaw's (2002) general conception of public deliberation we measured whether participants believed the ACP followed certain deliberative norms with the following questions: 1) How often do you believe participants just stated positions without justifying them?; 2) How often do you believe participants truly expressed what was on their mind?; 3) How often do you feel that other participants treated you with respect?; 4) When participants expressed views that were different from your own, how often did you consider what they had to say? These questions were answered on a 5-point scale from 1 (Never) to 5 (Almost always).

Benchmarking the Main Perspectives on Australian Democracy Using Q Methodology
We employed Q methodology (Brown, 1980;Stephenson, 1935) to measure participants' understandings of and preferences for a particular type of political system. We gathered statements from actual dialogue using a wide range of sources-including old and new media, World Cafés conducted by the New Democracy Foundation in the lead up to the ACP, and a report from the 2020 Summit (a gathering run by the Australian government in 2008). We drew a sample of 48 statements from this larger pool to comprise a manageable number for use in the Q sort at the ACP. These items were selected to encompass the broadest possible range of potential orientations toward Australian politics. Some examples are: We live in a great democracy; I don't want to be in a place where only a minority's viewpoint gets the right to say what's going to happen; The party system is the main obstruction to accountable politics. The set also included statements originally used in a prior study of Australian discourses of democracy (see Dryzek, 1994).
The Q sorts were obtained by asking participants, at each stage, to order the 48 statements into a set of eleven categories along a scale from -5 (Most disagree) to +5 (Most agree). In doing so, they assigned a score to each statement with the requirement that their ratings approximate a normal distribution (i.e., fewer statements can be placed in the extreme categories). Forty-eight individuals provided Q sorts at all four stages of the research, and it is these individuals that we examined in this study. To analyse how people understand the political system, we computed an inverted factor analysis. That is, we correlated the Q sorts provided by participants, as opposed to variables, using factor extraction (in this case Principal Components) to find the array of responses to statements (factor) that account for the highest possible variance among the Q sorts, and so on for the residual variation for the remaining factors. A subset of four of the resulting factors with the highest loadings were then transformed using varimax rotation to produce orthogonal positions that collectively accounted for most of the perspectives of participants (for a complete description of Q method see Brown, 1980).
Here the focus was on individual changes throughout the four stages of the deliberative process. Converging all the data from the four stages and disclosing the overarching structure allows the examination of how individuals changed their perspective in each of the four points in time.

Subgroup and Superordinate Identity
The ACP was a national representative assembly. As such, it mirrored the national parliament, which is comprised of members from the various electoral districts in Australia. Among other things, the role of these members is to represent the interests of the nation as well as of their electorate. This means that, as members of parliament, they are expected to navigate two social identities simultaneously-national and regional-and where the latter is included in the former. We therefore expected the same from the members of the ACP. Consequently, we regarded the ACP as a higher-order category that included the electoral subgroups.
Consistent with this reasoning, we measured subgroup identity in the form of identification with the ACP member's own home electorate. To do this we employed a single question: "To what extent did you represent the particular interests of your electorate?" The answer was provided on a five-point Likerttype of scale from 1 (Never) to 5 (Almost always). This measure was taken at the end of the deliberations.
Identification with the ACP (superordinate level) was measured with the question: "I see myself as an important part of the Citizens Parliament being held here in Canberra." This question was taken at the beginning of the deliberative process and similarly measured on a five-point Likert type of scale from 1 (Strongly disagree) to 5 (Strongly agree).

Was the Event Deliberative?
To make sure that the normative framework for deliberation had been followed, we first checked whether participants perceived deliberative norms in operation. The means in Table 1 indicate that participants in general perceived that the discussions followed basic deliberative norms, with each average score well above the midpoint. By way of comparison, a previous study juxtaposed these same data with equivalent measures taken regarding the Online Parliament that preceded the ACP (Gastil & Wilkerson, 2013). The ratings for both events were high, but the face-to-face ACP rated somewhat higher on both the rigor of its issue analysis and the democratic quality of the relations among its members. A separate analysis used independent coders to look at the transcripts of the ACP, and these ratings also gave favourable ratings on both dimensions-with the highest ratings coming for the social process-turn taking, listening, and respect (Gastil, 2013). Both these self-report survey ratings and transcript analyses are comparable to those obtained for an institutionally authorized minipublic in the Western United States-the Oregon Citizens' Initiative Review (Knobloch, Gastil, Reedy, & Cramer Walsh, 2013).
Finally, a study of the ACP transcripts found that equality of speaking rates among participants was high relative to comparison groups and comparable to another signal deliberative event, the 1996 National Issues Convention held in the United States (Bonito, Meyers, Gastil, & Ervin, 2013) Using a measure that assesses the variation in speaking rates within the breakout discussion tables at the ACP, the vast majority of table discussions rated as having "moderate equality" or higher. Another analysis of ACP speaking rates broken down by day showed quieter participants tending to speak up more by the fourth day, with many of those initially more vocal stepping back (Bonito, Gastil, Ervin, & Meyers, 2014).

What Understandings of Democracy Emerged?
To extract the narratives (understandings) of the Australian democratic system we conducted inverted factor analysis. Four factors emerged that contain the main narratives. These are named Narrative A, Narrative B, Narrative C, and Narrative D.

Narrative A: Liberal Democratic Contentment
Our democracy, although not perfect, is meaningful and works well, as governments do share power with citizens. We have a well-functioning party system, in keeping with the disagreement that is a necessary part of democracy. So, there is no need to be cynical about politics. The only slight problem is that money is too influential. Our system is able to change in the face of any problems that do arise. Citizens have a responsibility to pay attention and vote, and should be able to exercise influence.

Narrative B: Moralistic Leadership
While our system works reasonably well, we deserve better government than what we currently have. While we are all equal when we vote, money is too influential in politics. The key to better government is leadership that puts into practice moral principles such as those found in the bible. A robust democracy would feature decency, active citizenship, debate across different views, all motivated by what individuals think is in the public interest, not their private interests.

Narrative C: Anxious Majoritarianism.
While we live in a democracy, it is distorted by the influence of the rich, corporations, vested interests, and the power of various minorities who have too big a say. There is no reason for the interests of minorities and women to be promoted. Informed voting is really important: this is where the majority can express itself, and there is no need for citizens to be active beyond voting.

Narrative D: Discontented Participationists
Our political system is remote, closed, and inflexible. We are not well served by a conflictual party system that serves itself rather than the voters, where corporate lobbyists and vested interests have too much say. Politicians don't listen at all between elections. We deserve much better government than what we have. It would be great if we could vote on specific issues, not just on party platforms. Politics should involve much more in the way of citizen participation so that legitimate disagreements can be discussed, and policies can be influenced.

Longitudinal Changes in Understandings of Democracy
To track changes in understandings of democracy across the four stages we first analysed the average loadings for each factor to determine patterns among them. As displayed in Figure 2, all factors had relatively low mean loadings at stage one (rs between .17 and .24) with participants loading similarly on Narratives A and D. This pattern changed however markedly from stage one to stage four, with narrative A emerging as the strongest at stages two and three and culminating at stage four. Because factor loadings for all narratives, except for Narrative A, across time were lower than the conventional cut-off point of .4, the following analyses focus on Narrative A only. We conducted repeated measures ANOVA 3 to analyse changes in Narrative A. Pairwise comparisons with Bonferroni adjustment showed a significant difference in factor loadings between stage one and stages two (p = .01) and four (p = .001). Significant differences were also found between stages two and four (p = .001) and stages three and four (p = .001). The means displayed in Figure 2 show that there was a significant increase in factor loadings from stage one to stage two. No changes occurred from stage two to three, but loadings increased again significantly from stage three to four. In other words, there was a significant change in understandings of democracy from pre-to postdeliberation suggesting that deliberation had an effect on participants' understandings of "a strong democratic system." 3 Mauchly's tests showed that the assumption of sphericity was not violated, χ 2 (5) = 4.18, p = .52. The multivariate test using Wilks' statistic showed a significant difference in factor loadings across stages, λ = .35, F(3, 45) = 27.59, p = .001, pη 2 = .65. This result was further supported by within-subjects tests using Greenhouse-Geisser correction, F(2.83, 133.12) = 35.09, p = .001, pη 2 = .43.

The Relationship between Understandings of Democracy and Social Identification
To examine the relationship between subgroup and higher-order group identification with loadings on Narrative A at the end of the deliberations we conducted a hierarchical regression analysis. We also entered participants' loadings on this narrative at stage one to control for its effect on loadings at stage four. The mean for identification with the ACP was, M = 4.05 (SD = .75), well above the midpoint, and identification with the electorate subgroup was, M = 2.50 (SD = 1.2), right at the midpoint. These means indicate that participants identified more with the ACP than with their own electorate, suggesting that the higher-order identity was more salient than the subgroup identity. Moreover, the two group identifications were not significantly correlated with each other, r(86) = -.09, p = .44, suggesting independence from one another.

Table 2 Statistics Yielded from Hierarchical Regression Analysis Predicting Loadings on Narrative A at Stage Four.
Predictor variables zeroorder r Step We entered loadings on Narrative A-stage-one at step one, identification with own electorate (subgroup) at step two, and identification with the ACP (higherorder group) at step three. The results displayed in Table 2 show that both subgroup and higher-order identification predicted loadings on Narrative A, with higher-order identification adding variance over and above identification with the subgroup. Participants' position on Narrative A-stage-one did not contribute significantly to explaining their position on this narrative at stage four. The F-values show that preferences for Narrative A at the end of the deliberations are best explained by a model that includes subgroup and higherorder group identification. The zero-order correlations show that the more participants identified with the high-order group the more they loaded on Narrative A, whereas the more participants identified with the subgroup the less they tended to load on this narrative.

Discussion
Drawing on research on dual identity Saguy et al., 2007;Wenzel et al., 2009) this study examined how understandings of democracy change as a result of deliberation and how they are affected by identification with groups that are more or less inclusive. We found that participants tended to prefer a narrative of Liberal Democratic Contentment at all stages and this preference was particularly strong at the end of the deliberative process. We also found that this inclination was negatively associated with subgroup identification (electorate) but positively associated with higher-order group identification (nation).
It should be noted that all narratives contained a critical element of the system. However, Narrative A was the least critical of the four and the one that increased in loading, whereas the others remained relatively unchanged. Moreover, narratives A and D (Discontented Participationists) can be seen as opposites in terms of contentment versus discontentment. These two also had relatively similar loadings at the beginning of the ACP. From a perspective of group polarisation one could have expected both of them to have increased. That is, if positions were relatively polarised from the beginning, group discussions could have led to further polarisation. However, only Narrative A increased. This speaks against the idea that deliberation leads to group polarisation (Sunstein, 2000(Sunstein, , 2002 and that the deliberative norms in place are likely to prevent such dynamics from gaining ground.
It could arguably be said that the increase in contentment is a form of groupthink; that participants converged towards agreement about the democratic system. We argue that this is not the case because, although Narrative A got the highest loadings, these loadings were mid-range. In addition, preferences were distributed across all the narratives. This suggests that there was a diversity of preferences. More importantly, it speaks in favour of the idea that deliberation worked as expected as participants seem to have reached a greater understanding of the system that is both accepting and critical. In these days of populist politics and extreme beliefs that serve to undermine democracy, the current findings suggest that deliberative forums may help to counteract the populist tide.
We also argue that the greater level of satisfaction, rather than indicating groupthink, suggests that true deliberation took place, and that the preference for narrative A indicated critical contentment rather than mere contentment. We argue that being part of the ACP provided participants with a unique experience of democracy at work. The hallmarks of the forum such as the location in which it was held (i.e., Old Parliament House), the high profile of the speakers (i.e., chaired and attended by high-level public and political figures) and, not the least, the awareness that the final recommendations would be delivered to the Prime Minister, have certainly contributed to the belief that democracy can work.
These features are also likely to have enhanced identification with the ACP and to have made this identity more salient (Turner et al., 1987) which, consequently, explains why this identification was higher (as shown by means above the midpoint) and, therefore, more psychologically active, compared to subgroup identification. However, the data also indicate that subgroup identity was not switched off (mean at the midpoint). In fact, the simultaneous (although unequal) operation of both identities may be the mechanism that prevents some of the pitfalls of group dynamics (i.e., group polarisation, groupthink). These are ideas that need testing in future research. In addition, there are many other subgroup identities that can be relevant in deliberative contexts (e.g., gender, ethnicity, sexual orientation, etc.) and that are worth taking into account to examine how the interplay of superordinate/subgroup identities affects deliberative outcomes. Here it is important to note that, in any deliberative context, the subgroup identities investigated need to be relevant to the issue under discussion.
Further to the interplay between levels of identity, it is interesting to note that whereas higher-order identification was positively associated with Narrative A, subgroup identification was negatively associated with this position. This suggests that strong identification with the ACP was associated with stronger contentment with the system whereas stronger identification with one's regional base was associated with less contentment with the system. A possible explanation for this asymmetry could be that participants, who strongly identified with their region, felt more disconnected from Federal politics and therefore less satisfied with the system. To our knowledge, there is no available data from Australia analysing the association between satisfaction with the democratic system and identity. However, analyses (unpublished) of data from the European Social Survey Round 8 (ESS, 2016) show that the correlation (r = .26) between attachment to Europe (more inclusive identity) and satisfaction with the way democracy works in country is significantly higher (p < .001) than the correlation (r = .18) between the latter variable and attachment to country (less inclusive identity). These data provide an indirect indication of the adequacy of our reasoning.
Another important factor that we suspect also led to satisfaction with the system, is the deliberative question itself: "How can we strengthen the Australian democratic system to serve us better?" There is a positive charge in this formulation that implies that the system already is good. Participants may have reasoned that if the system is good why should we change it? Thus, all these factors together are likely to have led participants to develop a strong identification with the extant democratic system and prevented them from maintaining a stronger subgroup identity.
In important ways these findings advance theory and practice of deliberative democracy. First it informs deliberative theory by demonstrating that social identities at different levels of inclusivity are an important component of the deliberative process and can shape its outcomes. Specifically, allowing for both individual as well as social identity to be salient in the deliberative process is likely to prevent problematic group dynamics such as groupthink or group polarisation. As such, the psychological implications of group dynamics and how they interfere with/contribute to public reasoning need to be taken into account in the design of any deliberative forum. This insight further substantiates critical positions (e.g., Chambers, 2009;Dryzek, 2010) of deliberation as a rational form of discussion. Second, our results suggest that taking explicit account of the various levels at which participants identify with specific groups and the identities that may be particularly relevant for the issue at hand, inform the design of deliberative forums and offer a promising avenue to further advance knowledge and understanding of the dynamics at work in these settings. This is pertinent because group identification affects group dynamics regardless of whether it is accounted for in analyses. Failure to incorporate the measurement of social identity in the design of a deliberative event means that a great portion of the outcomes will go unexplained. Moreover, by incorporating identity as a variable, it enables the researcher to examine when and what evokes social versus personal identity.
A potential limitation of our study involves reliance on a small sample with results that can arguably be interpreted in different ways and prevent us from drawing more robust conclusions. Another limitation is the use of a single-item measure of identification, although there is evidence that reliable results are possible with as little as a single, overarching item to measure this construct (Postmes, Haslam, & Jans, 2012). Notwithstanding, in order to draw more conclusive statements about the role of social identification in deliberative processes further research into this domain with well-developed and validated measures is desirable.
Limitations aside, we believe that this investigation provides some lessons to be learned with regards to the design of deliberative events and the achievement of salient dual identity. In line with empirical work derived from the social identity perspective , dual identity allows for the presence of a critical stance that may emerge from the subgroup identity (particularly if this group is of a relative lower status) and at the same time the endorsement of a more inclusive group membership that contains goals that are common across subgroups. Deliberative forums need to be designed and implemented carefully to discourage over-identification with the deliberative group, promote dual, or multiple, identifications to engender a more critical position, in line with the deliberative endeavour. A potential avenue is to incorporate, in a mindful manner, social identity issues in the deliberative ground rules so as to capitalise on its potential and minimise the pitfalls.