The Virtues of Online Deliberation: An Experimental Study of the Impact of Facilitated Deliberation in Online Discussions among Like-minded Citizens.

Transcription

1 The Virtues of Online Deliberation: An Experimental Study of the Impact of Facilitated Deliberation in Online Discussions among Like-minded Citizens. Kim Strandberg, Kimmo Grönlund & Staffan Himmelroos Social science research Institute, Åbo Akademi University, Finland. Abstract According to deliberative theory, discussions with facilitation and deliberative rules (facilitated deliberation) should have more beneficial outcomes than discussions lacking facilitated deliberation. In the facilitated deliberation setting, group opinion polarization is expected to be alleviated and deliberators are expected to increase their knowledge, feelings of internal and external efficacy, levels of institutional and generalized trust as well as propensity for political action. We designed an experiment (N = 66) to test the extent to which introducing facilitated deliberation in online discussions among like-minded citizens has the aforementioned effects. Since both the like-minded setting and online discussions are often found to be lacking in deliberative quality, we perceive the experiment as a rather critical test of the outcomes of facilitated deliberation. Based on their attitudes toward a linguistic minority, participants were divided into a positive and a negative opinion enclave. Within the two enclaves, the participants were then randomly assigned to group discussions either using or lacking facilitated deliberation. We found that facilitated deliberation did alleviate group opinion polarization and also had some positive effects on the participants feelings of internal and external efficacy. Other positive effects where, however, not found due to facilitated deliberation. Introduction Modern democratic societies are increasingly characterized by a plurality of opinions and interests, and democratic theorists have emphasized the importance of including and exchanging different viewpoints as a part of democratic decision-making (cf. Cohen 1989). However, such exchanges are often not materializing in reality. Instead, there is an abundance of situations where only like-minded persons meet and discuss political issues in what has often been called enclaves. Such discussions among like-minded peers are generally considered to be problematic from a democratic viewpoint, and many Western democracies are indeed currently affected by a negative and sometimes even antagonistic discussion climate. Sunstein (2002; 1

2 2009) uses the term group polarization when describing this tendency of extremism and further remarks that it is especially a problem for deliberating groups among like-minded citizens, and often in discussions on the internet. Thus, in discussions on an issue which the participants initially agree on, group opinions tend to become more extreme as a result of the discussion. Other scholars who have studied online citizen discussions have also pointed to negative, antagonistic and generally poor quality discussions prevailing in online discussions (e.g. Jankowski & van Os 2004; Papacharissi 2004). However, there are several studies of deliberative mini-publics with small-n group discussion which show that opinions actually depolarize, or at least do not polarize, in a deliberative setting (e.g. Farrar et al. 2009, 2009; Himmelroos & Christensen 2014; Strandberg et al. 2016). Likewise, it has also been demonstrated that there is generally a better discussion climate in deliberations, both online and off-line, than in many other types of discussions (see Himmelroos 2011; Janssen & Kies 2005; Strandberg & Berg 2015). Many studies have also shown that deliberations tend to have an array of democratically beneficial outcomes for the participating citizens. These are, among others, changes in values and preferences concerning the issue of deliberation, knowledge gains on the issue, increased internal and external efficacy, generalized trust and trust for political institutions (Fishkin 2009; Fung 2003; Grönlund et al. 2010; Strandberg 2015). Why, then, might these effects occur due to deliberation and not in other discussions? One important difference between general discussions and deliberative mini-publics merits special attention here. In deliberative mini-publics, and other structured deliberative forums, the organizers improve the communicative process through the use of discussion rules and by having trained facilitators to uphold these rules (Moore 2012). According to Smith (2009, 169, 198), active facilitation is a central element of deliberation and is vital to ensuring that marginalized voices are heard. Although the importance of facilitation is often emphasized (Landwehr 2014; Moore 2012), research on the effects of facilitation is surprisingly scarce. Nonetheless, the studies which have been carried out on deliberative mini-publics i.e. without testing the effect of facilitated deliberation per se do indicate several positive outcomes stemming from deliberation (see Esterling et al. 2011; Fishkin 2009; Grönlund et al. 2010; Strandberg 2015). In this paper we aim to further investigate the effects of facilitated deliberation in an online enclave setting. We ask the general research question of whether enclave discussions with rules and facilitation bring about more beneficial effects than enclave discussions lacking 2

3 rules and facilitation. By focusing on enclave deliberations we would argue (see Coleman & Moss 2012, 6; Strandberg & Berg 2015, 165 for similar reasoning) that we are able to put facilitated deliberation to a harder-than-usual test since discussions among like-minded peers are far from the ideal speech situation depicted in the literature. For this purpose, we designed an online mini-public experiment (N=66) with small-n discussions in groups of like-minded citizens. The central focus point was the rules of discussion in conjunction with facilitation, what we choose here to call facilitated deliberation. Specifically, half of the discussions were carried out without rules and facilitation whereas the other half of the discussions used rules and facilitation. To achieve like-minded groups, the participating citizens were initially sorted into enclaves based on their opinion on the discussion topic and thereafter randomly assigned to the two discussion treatments within their respective enclaves. We will elaborate on the experimental design and procedures of the experiment later on in the paper. Before that section, though, the paper departs with presenting a theoretical backdrop for the study. In the backdrop we discuss themes such as enclave deliberations and their consequences, facilitated deliberation and the potential democratic outcomes associated with it and, finally, the online context as deliberative venue. Theoretical backdrop Enclave Deliberation Sunstein (2002, 2009) claims that there are a number of risks and problems related to political deliberation in like-minded groups, often referred to as enclave deliberation. He even raises questions concerning the future of democracy if people only speak and listen to others with like-minded views. In fact, we have now even witnessed acts of violence committed by people who had engaged in online discussion among their like-minded peers. The perpetrator of the 2011 Utoya massacre in Norway, for instance, had been engaged in online discussions with fellow right-wing extremists for many years prior to committing the act of terror (Bjurwald 2013). The fact that the online realm tends to attract people with similar opinions to interact with each other is a real concern in contemporary societies. What Sunstein basically points out as the root of the problem with enclaves is that discussion among like-minded tends to give rise to processes of group-think which usually have several adverse effects. He has first-andforemost highlighted group polarization of opinions (Sunstein 2009, 3), which means that groups of people with clear preconceived inclinations regarding an issue are likely to come out of a discussion on the issue with even more extreme views. This process has been argued in the 3

4 literature to have several potential causes pertaining to information biases and cognitive errors, social comparison and argumentation (Farrar et al. 2009; Isenberg 1986; Sunstein 2002). Social psychology literature shows that people are usually goal-directed when processing information and they often evaluate information with a directional bias toward reinforcing their pre-existing views (Nyhan & Reifler 2010). Thus, pre-existing views are likely to be strengthened in like-minded groups, because individuals tend to place higher merit in arguments supporting their own prior opinion (Mercier & Landemore 2012, 251; Wojcieszak 2011). Second, in a group where there is little or no alternative information or arguments available, individual biases in information processing and reasoning are not checked by arguments put forward by individuals supporting conflicting views. Sunstein argues that this can lead to informational cascades in the sense that misconceptions are amplified. Thus, opinion polarization might take place because the discussants mostly hear information and argumentation supporting their pre-existing position (Sunstein 2009, 22; Vinokur & Burnstein 1978, 873). Finally, it is also argued that the adverse effects of the like-minded setting could be due to peoples tendency to seek social acceptance and to adjust their behaviour according to what they see as the dominant position in the group (Brauer et al. 1995; Sunstein 2009, 23). Given that this paper is set in an online context, we would also draw some brief attention to other malaises found in online discussions. Although these may not be entirely only due to online discussion taking place in enclaves many have regarded the common online practice of discussing anonymously or under pseudonyms as an equally likely cause (Samuel 2004) they are certainly a manifest element of the polarization process occurring in said enclaves. We are, of course, referring here to poor discussion quality which been a major concern with online discussions. Thus, elements such as flaming, rudeness and less thoughtful contributions to the debate have been common findings in many studies of discursive quality in online discussions (e.g. Jankowski & van Os 2004; Papacharissi 2004; Witschge 2004). Facilitated deliberation In deliberative forums, which are designed to produce a high quality exchange of arguments, measures have been taken to battle the malaises of generic everyday discussions. Guiding these measures is the notion of an ideal deliberative process which consists of argumentation and fact-based discussion between free and equal citizens (Fung 2003, 344). This process should generally display several characteristics: inclusion, plurality of views, equality of discussion, 4

5 reciprocity, reasoned justification, reflection and sincerity. Of special relevance here are how organizers of deliberative forums seek to lessen the effects of group think and other harmful group mechanisms by incorporating various precautions, such as trained facilitators, rules for discussion, and unbiased background materials (Setälä, et al. 2010; Sunstein 2009, 56). Most often, these seem to have a desired effect in as much as scholars have not readily found traces of neither polarization of opinions nor negative discussion quality in deliberative settings (Farrar et al. 2009; Himmelroos 2011; Strandberg & Berg 2015; Strandberg 2015). With the term facilitated deliberation, we refer here to both the rules for deliberative interaction and the moderator or facilitator. Both of these are intertwined and central elements of a deliberative process. The former of these entails that participants are instructed to speak out, to listen to others, to behave respectfully, to learn about the issues and alternatives they are dealing with, and to try to persuade others of their points of view through spelling out reasons (Landwehr 2014). The latter, respectively, is tasked with upholding the rules of deliberation and with facilitating a good deliberative process. Facilitators assist discussion among the participants and try to ensure the satisfaction of forum participants (Mansbridge et al. 2006). They play a crucial role in breeding a discussion climate of mutual respect and reciprocity amongst the participants (Smith 2009). Pertaining to like-minded groups, the importance of facilitated deliberation is arguably augmented since the group composition per se does not ensure that alternative information nor counterarguments are presented. More concretely, the task of facilitated deliberation is then to make the participants critically examine the validity of their own arguments, to reflect on the fact that there are others (not necessarily present in the discussion) who hold views different from their own and even consider as to why this is so. Here the presence of discussion rules and facilitators becomes an instrument that may aid likeminded groups in overcoming groupthink and thus also avoid the adverse effects stemming from group think such as opinion polarization. Democratic outcomes of deliberation Deliberative mini-publics, or similar organized forums, have been argued to have several potential outcome-related effects. Basically, the common denominator for these outcomes is the idea of deliberation 'empowering citizens' in different ways (Fishkin 2009, 23-26) and enhancing their capacity for democratic participation (Fung 2003). The most often cited perceived effect of deliberation pertains directly to opinion polarization and concerns how 5

6 individual preferences and values are expected to change towards a more collective understanding through the deliberative process (Fishkin 2009; Grönlund et al. 2015). This could be labelled as the main democratic outcome of deliberative discussion. However, deliberation may also result in other side-effects (Grönlund et al. 2010). These are an enhancement of certain civic skills due to deliberation. The list could be made extensive but some of the commonly studied potential individual-level outcomes of deliberation are an increase in political knowledge, feelings of internal and external political efficacy, interpersonal trust and trust for political institutions (e.g. Fung 2003, ; Grönlund et al. 2010). Some studies (e.g. Fishkin et al. 2000; Grönlund et al. 2010) have indicated that participation in deliberation increases knowledge on the issue being deliberated albeit that it is sometime unclear precisely which element in the process causes this increase. Some (e.g. Muhlberger 2005) mainly point to the information package as a driver of knowledge gains (c.f. Grönlund et al. 2010). Nonetheless, that participants gain knowledge on the precise issue of discussion is perhaps an expected outcome, but Esterling et al. (2011, 485) also add that deliberators appear to 'encode policy information from outside the context of the deliberation', and achieve an increased general political learning. Moving on to internal political efficacy i.e. the participants confidence in their own political competence Gastil (2000, 358) claims that this is likely to be enhanced by taking part in a deliberation. Mutz (2006, 119), however, also points out an opposing view that being challenged and confronted with opposing views may cause doubts in one s position and beliefs. Grönlund et al. (2010, 98) indeed remark that the evidence on the effects of deliberation on internal efficacy is yet ambiguous. Regarding citizens feelings that their views and preferences have an influence on the political process, i.e. external efficacy some scholarly evidence exists in support of the notion that it might increase due to deliberation (Grönlund et al. 2010, 98; Strandberg 2015). Again, though, being confronted with differing views, and having one s own views challenged by others, may actually decrease the feeling of external efficacy as well. Fung (2003, 350) has also indicted that the impact of deliberation on external political efficacy might be partially dependent on whether the deliberation has an impact on actual policy or not. Interpersonal trust, or generalized trust, could be enhanced since the facilitated deliberation setting applies rules of sincerity and mutual respect (Dryzek and List 2003; Grönlund et al. 2010; Landwehr 2014). Likewise, it is logical to expect that trust in political institutions might be positively affected by participation in deliberation. Grönlund et al. (2010, 99) argue that especially trust for parliaments and governments i.e. legislative political 6

7 institution would be the most likely to increase among deliberators as they become aware of the complexities related to political issues and the political process. On the other hand, as Brown (2014, 59) for instance points out, citizens usually do not have time to learn these complexities and may therefore initially have blind trust in political institutions and experts. If this is the starting point before deliberation, it could actually be expected that deliberators decrease their trust in political institutions when forced to take part in a political process of sorts. Deliberation in the online context On the one hand, critical voices have been raised regarding the suitability of the online realm for citizens deliberations. Sunstein (2009) acknowledges that the internet has made it easier than ever before to come in contact with like-minded individuals and, consequently, opinion polarization seems perhaps to be most prominent online. We have also highlighted earlier in this paper how studies have found a less-than-ideal discussion climate online (e.g. Jankowski & van Os 2004; Papacharissi 2004; Witschge 2004). One the other hand, though, it has also been argued that the online context could be very suitable for deliberation if the venues per se are set up more-or-less to employ facilitated deliberation instead of being left to their own devices (Manosevitch 2010; Wright & Street 2007). The potential of online deliberation, as it has been discussed in the literature, seems to be due to logistical features as well as alleviation of certain adverse psycho-sociological factors. Regarding logistics, firstly, the online environment reduces the relevance of time, access, and geographical distances which are often significant obstacles in organizing citizen deliberations such as mini-publics offline (see Manosevitch 2010). Secondly, some scholars have argued that the online environment might actually free people from psychological barriers which often have restricted their engagement in off-line deliberation. Stromer-Galley (2002, 35-38) found that the feeling of being private when sitting by your computer at home while engaging in online political conversations provides a relief for citizens who might feel uncomfortable in face-to-face settings. Moreover, scholars have argued that the online context could reduce the influence of status and social position in discussions and in deliberation (e.g. Gastil 2000, 359). In short, the core of this line of argumentation is that focus shifts from who is saying something to what is actually said which would make the 'force of the better argument' likelier to prevail. 7

8 Taking a brief look, finally, at findings from studies of online deliberation, the first thing to notice is that online deliberation has generally been found to work in similar manner as offline deliberations (e.g. Grönlund et al. 2009; Luskin et al. 2006). More specifically, the main focus has often been on opinion changes and knowledge gains. Regarding these, Luskin et al. (2006) show that deliberators changed their attitudes and that deliberating reduced variance in opinions (see also Grönlund et al. 2009; Muhlberger 2005; Price & Capella 2002; Smith et al. 2009; Strandberg 2015; Triantafillidou et al. 2015). Strandberg et al. (2016) explicitly tested opinion polarization among groups of like-minded citizens and found that facilitated deliberation alleviated group polarization. Learning, or knowledge gains, are also often reported. Price and Capella (2002, ), found that deliberation increased participants knowledge specifically on certain issues of discussion. Luskin et al. (2006) too report small but nonetheless statistically significant knowledge increases as do others (Grönlund et al. 2009; Min 2007; Strandberg & Grönlund 2012). Scholars have also considered the side-effects of online deliberation to some extent. Min (2007) found that taking part in an online deliberation significantly increased both participants efficacy and willingness to participate in politics. Price and Capella (2002, ) also state that deliberators had become significantly more willing to engage in politics and other civic activities after taking part in their long-term deliberation project and that discussion attendance lead to increased generalized trust. Strandberg (2015) also found that deliberating online positively affected citizens feelings of efficacy and their trust for political institutions. Hypotheses Essentially, three central elements can be traced in our theoretical backdrop. One is that enclave deliberation, especially in the online context, tends to have mainly adverse effects if left unchecked by procedural safeguards. We have mainly highlighted group opinion polarization and sub-par discussion quality here. The other element basically states that these malaises can be alleviated, if not remedied entirely, if online discussions are designed to more closely mimic the ideals of deliberative theory often through applying both discussion rules and a facilitator. Finally, we have presented the bulk of research in which various outcome-related effects are predicted to occur due to deliberation, both off-line and online. Given that the purpose of this study is to test whether enclave discussions with rules and facilitation bring about more beneficial effects than discussions lacking rules and facilitation, an array of hypotheses emanate. Thus, we expect facilitated online deliberation in like-minded groups to: 8

9 H1: alleviate group opinion polarization tendencies, whereas lack of facilitated deliberation leads to opinion polarization. H2: result in higher levels of knowledge among participants than in discussions lacking facilitated deliberation. H3: result in higher internal and external efficacy among participants than in discussions lacking facilitated deliberation. H4 lead to a greater level of generalized trust among participants than in discussions lacking facilitated deliberation. H5: result in greater levels of trust for political institutions among participants than in discussions lacking facilitated deliberation. Experimental procedure and design The procedure of the experiment The deliberation was designed as a Lab-in-the-field experiment in which the purpose was to compare the outcomes of deliberation between like-minded groups. Some groups were subject to a treatment of having facilitated deliberation, whereas the rest discussed freely without rules. The use of a Lab-in-the-field strategy is often seen as an efficient trade-off between the benefits of a controlled lab experiment, such as a high internal validity, and the higher level of external validity of the field experiment (Morton & Williams 2010, ; Strandberg 2015, 12). The topic for deliberation used in the experiment was the status of the Swedish language in Finland. The Swedish language is an official language in Finland and at the time of the experiment this topic was very salient due to an approaching vote in the Parliament on a citizens initiative 1, according to which the teaching of Swedish in schools would have been made voluntary. The experiment started with the formation of groups of like-minded citizens, henceforth called opinion enclaves. This was done by the aid of an initial survey which was sent out to a simple random sample of citizens. This initial survey, which was disseminated 15,000 adult citizens (ages years) at the end of September 2014, measured opinions on the Swedish 1 The Finnish Citizen Initiative institution was implemented through an amendment to the Constitution of Finland in According to this, an initiative requires the signatures of 50,000 eligible voters so that the Parliament will process it. The institution is indirect since there is no referendum and the final decision is made by the Parliament. 9

10 language using 17 items (this survey is henceforth called T1). The survey had been developed by a broader team of scholars and also tested in a pilot study with university students, in order to assure the validity of the questionnaire items. In total, 1,509 persons responded to the T1 survey. Through factor analysis, we thereafter established that 14 of the 17 items loaded on one factor 2. These 14 items were then used to construct a sum variable ranging from 0-14, according to which a higher score on the scale corresponds to a more positive opinion on the Swedish language. The dispersion of attitudes on the summarized scale can be found in Appendix A. In order to form the actual enclaves, respondents with a summarized value of less than 6.01 were grouped into a con-enclave of N = 563 (negative attitudes toward the Swedish language) and respondents with a score greater than 8.0 into a pro-enclave of N = 697 (positive attitudes). The remaining respondents (N = 249), i.e. those who were neither clearly in favour of nor against the Swedish language, were excluded since the experiment was only about discussion among like-minded citizens. All the respondents who were part of either the con- or the pro-enclave and had expressed their interest in taking part in the deliberation were invited to take part. The participants were at this stage promised a compensation (a gift voucher worth 90 Euros) upon finishing the entire experiment. Of those invited, 110 participants eventually turned up for the online deliberation. It is noteworthy, though, that the entire experiment included additional treatments which are not relevant for the purposes of this paper. Thus, the analyzed sample here consists of N = 66 participants. Design of the experiment The experimental design of the deliberation is twofold. Firstly, a single-factor (the factor being facilitated deliberation or not) pre-test post-test experimental design was used for testing the effects on opinion polarization (Morton & Williams 2010). Secondly, a single-factor (again facilitated deliberation or not) post-test only design was employed for testing all the other outcome effects. This dual design was necessary since opinion polarization needs to be tested as a change measure pre- to post-test while, at the same time, the initial survey needed to be short in order to not discourage potential participants whereby we chose to ask questions about the remaining outcomes of deliberation only post-test. These designs were used within each opinion enclave so that participants within each enclave were randomly assigned to groups with 2 Cronbach s alpha of.94 (see Appendix B for items). 10

11 facilitated deliberation and to groups lacking rules and facilitation. Table 1 shows the allocation of participants to treatments 3 : [Table 1 about here] Even though the random allocation to treatment groups usually takes care of equivalence between the groups, we chose to double-check the equivalence of the treatment groups since the number of participants in our experiment was low (see Gribbons & Herman, 1997 for discussion). Our data allowed us to do this check for five background variables (chi-squares test): gender, age, education, profession and region of residence. None of these displayed any significant differences between treatment groups (lowest p = 0.21), suggesting that the randomization had worked as intended and the treatment groups were equivalent to each other. The actual deliberations took place in November None of the participants were aware of the different experimental treatments before deliberating but, as is customary according to good ethical practice in experimental research, a debriefing material which explained the true purpose of the experiment was mailed to them after the experiment. Each participant only took part in one group discussion in one session. After logging in to our website for the first time, participants answered a survey measuring extensive background information (abbreviated to T2). Each participant also read a four-page leaflet, which moreover served as an initial basis for discussion, containing unbiased information on the Swedish language in Finland. After the discussion had ended, the participants answered a post-test survey (T3) which contained, among other items measuring the outcomes of deliberation, the same items on the opinions on the Swedish language which had been included in the pre-test survey (T1) as well. Each discussion group consisted of around six participants and discussed for approximately one and a half hour. The discussions used live web-cam feeds as well as audio and text and half of the groups employed facilitated deliberation with rules (henceforth called the Rules treatment) and half of them had no rules (henceforth: the No Rules treatment). 3 The final number of participants varied somewhat between cells in each mode due to people not showing up for the actual deliberation. Using an online calculator: the final number of participants results in the following approximated statistical power at the 0.10-significance level: effect size = 0.65 (Cohen s d), statistical power =

12 Regardless of the treatment, though, a person was present in the discussion room. His/her role however varied greatly from being an active facilitator to a passive observer according to the treatment. Table 2 shows the main differences in operationalization between the Rules and No Rules treatments. [Table 2 about here] Dependent variables In our analyses, opinion polarization was measured in each enclave by the change T1 to T3 in the summarized opinion scale on the Swedish language (0-14 point scale), whereby a negative change-value in the con-enclave as well as a positive change-value in the pro-enclave would indicate polarization. We also provide a measure of absolute polarization which combines the pro and con opinion enclaves into one sample and measures how much polarization or depolarization occurred in each treatment. Absolute polarization is operationalized as increased anti-swedish opinions in the con enclave or increased pro Swedish opinions in the pro enclave after the experiment. Thus, positive values indicate polarization, and negative values depolarization. The other democratic outcomes of deliberation i.e. knowledge, internal- and external efficacy, generalized trust, trust for political institutions as well as propensity for political action were all measured at T3 using standard survey items which have been used previously in similar research (see Appendix C for full list of items). All of these were compared between treatment groups at T3 per the post-test only logic which would state that, since there is an initial randomization into treatments, all groups are equivalent at the outset and any observed post-test differences are thus logically due to receiving different treatments. Findings The presentation of the findings is done in the order of our hypotheses. Statistical significance is indicted at alpha= 0.10 in order to maintain a sufficient statistical power in the analyses and have a lower combined Type I and Type II error in the analyses 4. All comparisons are carried 4 Significance-level (alpha) at 0.05 would yield a statistical power of 0.80 (Kraemer & Thiemann 1987, 110) which combines into a total risk of 0.25, whereas alpha 0.10 at power 0.88 results in a total risk of

13 out using two-tailed independent samples t-tests. Since the N is low in the experiment, all of the findings presented have also been confirmed using non-parametric tests (Mann-Whitney Test of mean ranks). We start by addressing hypothesis H1: Facilitated deliberation alleviates group opinion polarization tendencies in online deliberation among like-minded citizens whereas lack of facilitated deliberation leads to opinion polarization. Table 3 presents the polarization tendencies, both per enclave and treatment as well as the absolute polarization. [Table 3 about here] Table 3 basically shows two things: firstly, there is one statistically significant trend found for the con-enclave according to which there is a difference in polarization between the Rules- and No Rules- treatments. Secondly, although the other tendencies are not statistically significant (the one for absolute polarization being the closest), they all point in the direction of Rules alleviating polarization and that a lack of rules results in polarization. These findings indicate that H1 is mostly, but not entirely, confirmed. In Table 4, presented next, we attend to hypotheses H2: Facilitated deliberation in online deliberation among like-minded citizens results in greater learning among participants than in discussions lacking facilitated deliberation and H3: Facilitated deliberation in online deliberation among like-minded citizens results in higher internal and external efficacy among participants than discussions lacking facilitated deliberation. [Table 4 about here] Judging from the findings in Table 4, we can dismiss H2 since there were no significant differences between the treatments concerning knowledge levels in either enclave. The difference in the con enclave was somewhat close to being a significant trend (p = 0.16) but that difference was contrary to the hypothesis since the No Rules treatment had higher level of knowledge after deliberation than did the Rules treatment. This finding could tentatively be due to our choice of providing the same information package to all participants indirectly leading 13

14 to them feeling a pressure to make use of it in their discussions, also when lacking facilitated deliberation. Turning to hypothesis H3, the findings provide some support for the hypothesis. There is one significant trend for internal efficacy (in the pro enclave) and one which is close to significant (in the con enclave). Both of these indicate that participants in the facilitated deliberation treatment had a comparably higher level of internal efficacy after deliberation. Additionally, there is one difference concerning external efficacy in the pro enclave which is almost significant (p = 0.12) and indicates that the Rules treatment has a higher external efficacy post-deliberation than the No Rules treatment. The analyses now proceed by looking, in Table 5, at generalized trust and trust for political institutions. To reiterate, this concerns hypotheses H4: Facilitated deliberation in online deliberation among like-minded citizens leads to a greater level of generalized trust among participants than discussions lacking facilitated deliberation as well as H5: Facilitated deliberation in online deliberation among like-minded citizens results in greater levels of trust for political institutions among participants than in discussions lacking facilitated deliberation. [Table 5 about here] There is an interesting side-observation to be made from Table 5 and that is the stark differences in the levels of trust between the two enclaves. Seemingly, the con-enclave has gathered people who are generally much more distrusting than the people in the pro-enclave. Moving on to the analyses of relevance though, we see that hypothesis H4 regarding generalized trust is not confirmed at all. The same hold true for hypothesis H5 concerning which the significant findings and one which is close to significance (all in the con-enclave) show higher levels of institutional trust for the No Rules treatment, not for the Rules treatment. Conclusions At the outset of this paper we discussed the potential dangers inherent in discussions among like-minded citizens, especially online. Moreover, we presented both argumentation and evidence in support of the notion that deliberative discussions ought to alleviate some of the problems with online enclave discussions. In hindsight, then, what are the major conclusion to 14

15 be made from our study regarding the impact of facilitated deliberation in enclave deliberations online? Generally, our findings show that enclave discussion with discussion rules and facilitation does bring about more beneficial effects than discussions lacking these procedural safeguards. Give that our experiment was just a one-off discussion session, thus making the noted effects likely to amplify upon repeated or long-term deliberation, these are encouraging findings in today s climate of hasty opinions and polarization even at the societal level. Furthermore, our findings provide guidance for policy-makers who wish to effectively engage citizens further in political processes. Thus, our experiment does add further proof concerning how opinion polarization is alleviated in like-minded settings when certain procedural conditions are met (c.f. Grönlund et al. 2015; Schkade et al. 2010, 243; Strandberg et al. 2016). We also demonstrate how citizens subjective feelings of both internal- and external efficacy received a boost due the facilitated deliberation treatment, even in online enclave settings (c.f. Strandberg 2015). However, and this is a rather big caveat, the effects of facilitated deliberation are by no means omnipotent. In the experiment reported on here, to illustrate this point, only two out of six hypotheses were confirmed. Beyond alleviating opinion polarization and boosting feelings of efficacy, thus, we found little proof of facilitated deliberation having positive effects. Sometimes, the effects even ran contrary to our expectations. For instance, the fact that we found no effects for generalised trust is interesting. Basically, the argument was that the deliberative setting would have positive effects on this since facilitated deliberation should breed a culture of mutual respect and civility. So how could we explain that there were no effects on generalised trust? Two answers seem plausible; 1) people need to take others into account even when discussing without facilitated deliberation, or 2) the like-minded setting means that the deliberators do not really encounter new types of people which might have made them become more trusting of the unknown. As to the findings running contrary to expectations i.e. institutional trust for which the lack of facilitated deliberation showed higher post-deliberation levels than the facilitated deliberation treatment we can only provide a very tentative interpretation. Perhaps this is in line with the perspective of Brown (2014, 59), which we discussed earlier, that citizens may initially have blind trust in political institutions and experts and could decrease their trust in political institutions when forced to take part in a political process and experience how complex this process may actually be. The facilitated deliberation treatment would thus be fostering critical citizens. 15

16 There are, naturally, limits to how far we can generalize from the findings of this one experiment. Firstly, although our ambition was to gather a large number of participants, the final number of participating citizens turned out to be low. This makes it hard to claim, with any real confidence, that our findings are valid for a larger population. It should be stressed here, though, that the purpose of factorial experiments is precisely to test the factors i.e. what happens to the outcomes if we vary this variable so the ambition to generalize mainly concerns the factors, not the participants. Given, also, that the like-minded setting is far from the most fruitful setting for finding positive effects due to deliberation, we would argue that our findings concerning the facilitated deliberation factor may be rather robust. Further experiments carried out in other settings and using larger numbers of participants would of course confirm our belief. A further limiting aspect is a slight uncertainty regarding internal validity in online experiments, which are often due to technical limitations. Thus, there might be confounding factors such as audio or video problems in the discussion room which may affect the findings to some degree. Moreover, even though it is often cited as one of the logistical benefits of online deliberation (e.g. Manosevitch 2010), the fact that people are not physically present in one location makes it harder to ensure that they conduct themselves in the way intended by the experimental design. One final critique which is also often aimed at studies of deliberative outcomes is the lack of focus on the actual process itself how can we be sure that the deliberative setting was, in fact, more deliberative than the non-deliberative setting? Although we have not conducted any content analyses of the actual discussion in the experiment either, our T3 post-test survey did contain questions about the participants subjective experience of the discussions. Post-hoc analyses of these questionnaire items (not reported here due to length reasons but available on request) show that participants in the facilitated deliberation discussion groups felt that the discussions had been more in line what we, as scholars, would identify as features e.g. lack of dominance by one participant, staying on topic, less disagreement etc. of a deliberative process. To conclude, within certain constraints, this study has served to demonstrate that facilitated deliberation at least does more good than harm in online discussions among likeminded citizens. It alleviates group opinion polarization and fosters empowered and critical citizens. The crucial challenge in the future will be how to scale-up facilitated deliberation to the societal-level so that such positive outcomes would benefit more than just participants in a scholarly experiment. 16

17 17

18 References Bjurwald, L. (2013). Skrivbordskrigarna: Hur Extrema Krafter Utnyttjar Internet [Keyboard Warriors: How Extreme Forces Make use of the Internet]. Stockholm: Natur & Kultur. Brauer, M., Judd, C. M., & Gliner, M. D. (1995). The Effects of Repeated Expressions on Attitude Polarization During Group Discussions. Journal of Personality and Social Psychology, 68(6), Brown, M. B. (2014). Expertise and Deliberative Democracy. In S. Elstub & P. McLaverty (Eds.), Deliberative Democracy. Issues and Cases (pp ), Edinburgh: Edinburg University Press. Cohen, J. (1989). Deliberation and Democratic Legitimacy. In D. Matravers & J. E. Pike (Eds.), Debates in Contemporary Political Philosophy: An anthology. Oxford: Routledge. Coleman, S. & Moss, G. (2012). Under construction: the field of online deliberation research. Journal of Information Technology & Politics 9, Dryzek, J. S. & List, C. (2003). Social choice theory and deliberative democracy: a reconciliation. British Journal of Political Science 33, Esterling, K., Neblo, M. & Lazer, D. (2011). Means, motive, an opportunity in becoming informed about politics: A deliberative field experiment with members of congress and their constituents. Public Opinion Quarterly 75(3), Farrar, C., Green, D., Green, J. E., Nickerson, D. W., & Shewfelt, S. (2009). Does discussion group composition affect policy preferences? Results from three randomized experiments. Political Psychology, 30(4). doi/ /j x Fishkin, J. S. (2009). Virtual public consultation: prospect for internet deliberative democracy. In T. Davies & S. P. Gangadharan (Eds.), Online Deliberation: Design, Research, and Practice (pp ), Stanford, CA: CSLI Publications. Fishkin, J. S., R.C. Luskin & Jowell, R. (2000). Deliberative polling and public consultation. Parliamentary Affairs 53, Fung, A. (2003). Survey Article: Recipes for Public Spheres: Eight Institutional Design Choices and Their Consequences. Journal of Political Philosophy, 11(3), doi: / Gastil, J. (2000). Is Face-to-Face Citizen Deliberation a Luxury or a Necessity?. Political Communication, 17(4), doi: / Gribbons, B., & Herman, J. (1997). True and Quasi-experimental Designs. Practical Assessment, Research & Evaluation, 14(5). Grönlund, K., Strandberg, K., & Himmelroos, S. (2009). The Challenge of Deliberative Democracy Online A Comparison of Face-to-Face and Virtual Experiments in Citizen Deliberation. Information Polity, 14(3), doi: /ip Grönlund, K., Setälä, M. & Herne, K. (2010). Deliberation and civic virtue lessons from a citizen deliberation experiment. European Political Science Review 2, Grönlund, K., Herne, K., & Setälä, M. (2015). Does Enclave Deliberation Polarize Opinions? Political Behavior. doi: /s x Himmelroos, S. (2011). Det Demokratiska Samtalet. En Studie av Deliberativ Demokrati i ett Medborgarforum [The Democratic Conversation. A Study of Deliberative Democracy in a citizens forum]. Doctoral thesis, Turku, Finland: Åbo Akademi University Press. Himmelroos, S., & Christensen, H. (2014). Deliberation and Opinion Change: Evidence from a Deliberative Mini public in Finland. Scandinavian Political Studies, 37(01), doi: / Isenberg, D. J. (1986). Group polarization: A critical review and meta-analysis. Journal of Personality and Social Psychology, 50(6),

19 Jankowski, N. W., & Os, R. G. van. (2004). Internet-based political discourse : a case study of electronic democracy in Hoogeveen. In P. M. Shane (Ed.), Democracy online : the prospects for political renewal through the Internet (pp ). London: Routledge. Janssen, D. & Kies, R. (2005). Online forums and deliberative democracy. Acta Politica 40(3):, Landwehr, C. (2014). Facilitating Deliberation: The Role of Impartial Intermediaries in Deliberative Mini- Publics. In K. Grönlund, A. Bächtiger, & M. Setälä (Eds.), Deliberative Mini-Publics: Involving Citizens in the Democratic Process (pp ). Colchester: ECPR Press. Luskin, R.C., Fishkin, J.E. & Iyengar, S. (2006). Considered opinions on U.S. foreign policy: evidence from online and face-to-face deliberative polling. California: The Center for Deliberative Democracy. Manosevitch, E. (2010). Mapping the practice of online deliberation. In F. De Cincio, A. Machintosh, & C. Peraboni (Eds.), Online Deliberation. Leeds, UK: Fourth International Conference. Mansbridge, J., Hartz-Karp, J., Amengual, M., & Gastil, J. (2006). Norms of Deliberation: An Inductive Study. Journal of Public Deliberation, 2(1), 7. Mercier, H., & Landemore, H. (2012). Reasoning is for arguing: Understanding the successes and failures of deliberation. Political Psychology, 33(2), doi: /j x Min, S.-J. (2007)., Online vs. face-to-face deliberation: effects on civic engagement. Journal of Computer Mediated Communication, 12, Moore, A. (2012). Following from the front: theorizing deliberative facilitation. Critical Policy Studies, 6(2), doi: / Morton, R. B., & Williams, K. C. (2010). Experimental Political Science and the Study of Causality: From Nature to the Lab. Cambridge: Cambridge University Press. Muhlberger, P. (2005).The Virtual Agora project: a research design for studying democratic deliberation. Journal of Public Deliberation 1(1). Mutz, D. C. (2006). Hearing the Other Side: Deliberative Versus Participatory Democracy, New York: Cambridge University Press. Nyhan, B., & Reifler, J. (2010). When Corrections Fail: The Persistence of Political Misperceptions. Political Behavior, 32(2), doi: /s Papacharissi, Z. (2004). Democracy online: civility, politeness, and the democratic potential of online political discussion groups. New Media & Society, 6(2), doi: / Price, V. & Capella, J. N. (2002). Online deliberation and its influence: the electronic dialogue project in campaign IT & Society 1(1), Samuel, A. (2004). Hacktivism and the Future of Political Participation. Doctoral thesis, Cambridge, MA; Harvard University Press. Schkade, D., Sunstein, C. R. & Hastie, R. (2010). When deliberation produces extremism. Critical Review: A Journal of Politics and Society, 22(2-3). Setälä, M., Grönlund, K., & Herne, K. (2010). Citizen deliberation on nuclear power: A comparison of two decision-making methods. Political Studies, 58, doi: /j x Smith, G. (2009). Democratic Innovations: Designing Institutions for Citizen Participation. Cambridge: Cambridge University Press. Smith, G., Sturgis, J. P. & Nomura, H. (2009). Deliberation and internet engagement: initial findings from a randomized controlled trial evaluating the impact of facilitated internet forums. Paper, ECPR General Conference, Potsdam, Germany. Strandberg, K. (2015). Designing for democracy?: an experimental study comparing the outcomes of citizen discussions in online forums with those of online discussions in a forum designed according to deliberative principles. European Political Science Review, 7(03), doi: /s

20 Strandberg, K. & Berg, J. (2015). Stretching the boundaries of online deliberation An experimental study evaluating the impact of temporality and identifiability on discussion quality in online deliberations. Javnost The Public 22(2), Strandberg, K., & Grönlund, K. (2012). Online Deliberation and Its Outcome Evidence from the Virtual Polity Experiment. Journal of Information Technology & Politics, 9(2), doi: / Strandberg, K., Himmelroos, S. & Grönlund, K. (2016). Can deliberative norms prevent group polarization in like-minded groups?. Paper, MPSA annual conference, Chicago. Stromer-Galley, J. (2002). New voices in the public sphere: a comparative analysis of interpersonal and online political talk. Javnost The Public 9(2), Sunstein, C. R. (2002). The Law of Group Polarization. Journal of Political Philosophy, 10(2), doi: / Sunstein, C. R. (2009). Going to Extremes: How Like Minds Unite and Divide. Oxford: Oxford University Press. Triantafillidou, A., Lappas, G., Yannas, P. & Kleftimodos, A. (2015). The role of online deliberation on citizens attitudes. International Scholarly and Scientific Research & Innovation 9(8), Vinokur, A., & Burnstein, E. (1978). Depolarization of attitudes in groups. Journal of Personality and Social Psychology, 36(8), doi: / Witschge, T. (2004). Online deliberation: possibilities of the internet for deliberative democracy. In P. Shane (ed.), Democracy Online: The Prospects for Political Renewal through the Internet (pp ), New York: Routledge. Wojcieszak, M. (2011). Deliberation and Attitude Polarization. Journal of Communication, 61(4), doi: /j x Wright, S., & Street, J. (2007). Democracy, Deliberation and Design: The Case of Online Discussion Forums. New Media & Society, 9(5), doi: /

21 Appendix A. Initial dispersion of attitudes among respondents to T1 survey (N = 1,509) 21