The emergence and expansion of digital technologies has attracted significant academic interest in terms of their impact on the public sphere and, more generally, on politics and democracy itself. Researchers have been divided from the early nineties into those who emphasized the potential of the Internet to improve democracy and revolutionize politics (Rheingold 2002; Lévy 2002; Jenkins 2006) and those who held rather prudent and skeptical stances: either predicting that digital technologies would just be normalized by the particular socio-political contexts in which they were deployed (Davis 2005; 2001; Margolis and Resnick 2000) or warning against the different threats to democracy that the Internet entailed (Sunstein 2001; Precht 2010).

On the one hand, cyber-optimists contended that the Internet would contribute to improving democracy by offering new channels of direct communication with political elites, increasing public transparency and allowing citizens to self-organize and participate in both conventional and non-conventional politics (Rheingold 2002; Lévy 2002; Jenkins 2006). Some authors even claimed the rebirth of direct democracy thanks to digital technologies, which would strengthen social capital, increase self-government practices and provide the material infrastructure for direct decision making (Lévy 2002).

This view of communication technologies as a driving force of political and social progress is nothing new. Technological optimism has emerged periodically after every innovation since the industrial revolution (Davis 2001; Vedel 2003). In that sense, cyber-optimists have been harshly criticized for their utopianism and their tendency to reduce the problems associated with democracy to their procedural aspects. For example, Morozov (2013; 2011) argued that cyber-optimists’ approaches either exhibit deep ignorance about the problems associated with political theory and democracy functioning; or they deliberately refuse to confront them with the required seriousness and rigor. In an article in News Republic, he wonders:

For their part, cyber-pessimists have shown skepticism about ICT’s capacity to consolidate democracy in a number of ways. Some highlighted the resistance to democratization of political elites and most citizens’ lack of interest in politics (Davis 2001). Some conceptualized the impact of the Internet as that of a simply symbolic revolution (Mazzoneli 2001), limited to esthetic aspects of political communication but unable to engender a transformation of political practices. Others argued that the revolutionary potential of the Internet would be normalized by the socio-political reality leading to politics as usual (Margolis and Resnick 2000).

A few scholars even warned against different dangers associated with ICT’s, such as communicative endogamy, the reduction of social empathy or the drastic suppression of exposure to diversity (Sunstein 2001; Prior 2007; Precht 2010). One of the classic critiques of digital communication’s impact on democracy harkens back to the work of Cass R. Sunstein (2007; 2001), who predicted the social isolation, ideological polarization and balkanization of the public sphere as a result of the renewed power of informative selection that users are provided with online. Sharing this rationale, Precht (2010) conceived the Internet as a perfect machine to intensify prejudices (p. 480), thus facilitating the natural human tendency to reinforce preexisting views of the world and to preserve cognitive internal consistency (Festinger 1957).

In sum, this theoretical schism has largely dominated scholarship on the Internet’s impact on politics during the last two decades, to such an extent that, according to Wright (2011), it has deeply harmed the empirical analysis of political conversation online in a number of ways. First, it has disproportionally influenced the specific aspects of the political use of the Internet that are actually analyzed. Studies generally address existing political actors’ and institutions’ Internet practices, while the interactions among ordinary citizens are rather ignored by academia (Jenkins and Thornburn 2003; Graham 2010; Wright 2011).

Second, it has predominantly led researchers to formulate narrow definitions of political discussion grounded in very strong normative standards, mostly in deliberative theory (Freelon 2010) that are easily refutable in empirical terms, since informal and non-regulated deliberative environments always tend to be fragmented and disperse (Martí 2006). In that sense, even skeptical scholars analyze online political conversations through operationalizing Habermasian notions of idealized deliberation (see, for example, Valera 2014a; 2012), which implies assuming part of the cyber-optimist’s revolutionary rhetoric of which they are so critical (Wright 2011).

Third, this schism has also driven scholars to interpret empirical results in a very negative way, as long as data do not meet the high normative standards (Wright 2011). In fact, researchers operationalizing ideal notions of deliberation constantly face the uncomfortable task of trying to answer whether the analyzed conversations are deliberative or not (Black et al. 2011), since there are no absolute threshold levels from which scholars can draw any decisive conclusions. They are therefore faced with questions like: a) how to establish concrete threshold levels that variables need to exceed in order to conceive discussions as deliberative; b) how to justify these threshold levels; and c) how to infer any theoretical knowledge about the democratic value of online political talk.

These limitations derived from operationalizing normative notions of deliberation and interpreting ensuing results have often led most researchers to basically confirm that online political discussions do not reflect either all or some of the deliberative norms (see, for example, Wilhelm 1998; Dumoulin 2002; Dalhgren 2005; 2000; Jensen 2003; Ruiz et al. 2010; Valera 2014a; 2012). This empirical rejection of the revolutionary scenario predicted by the cyber-optimist school (though valuable per se) does not really expand our understanding of how these online political discussions are taking place and which democratic functions they are serving (if any).

Moreover, in researching online political conversations through normative deliberative notions, we might be overlooking the actual democratic value of these spontaneous interactions among citizens (Coleman and Blumler 2009; Freelon 2010; Wright 2011). Indeed, the fact that some of the variables used to measure deliberation (such as discursive diversity, argumentation or reciprocity) score poorly does not necessarily mean that these spaces are not serving any democratic purposes. It just means that they do not meet the deliberative standards associated with Habermasian notions.

“The predominance of deliberation as an analytical framework has thus led to the relative neglect of online discussion characteristics not classified under its domain” (Freelon 2010:3). To the best of our knowledge, only a few scholars have moved beyond these formal notions of deliberation when analyzing online political talk to address understudied aspects like humor or emotions in online political dialogue (Graham 2010; Sampietro and Valera 2015). In that regard, Freelon (2010) contended that deliberation is not the only democratically valuable type of online political discussions, and suggests taking into account other analytical frameworks of democratic communication grounded in the liberal individualist and the communitarian traditions.

So far, scholarship in online political discussions has generated some shared knowledge about the factors that influence the deliberative quality of conversations. Researchers agree that the quality of online political conversations partially depends on the design of the discussion space where they take place (Beierle 2004; Noveck 2004; Wright 2005; Stromer-Galley and Wichowski 2011; Black et al. 2011). For example, in their study about European Union and UK online forums, Wright and Street (2007) argued that web and interface design constitute relevant factors influencing the possibility of deliberation. These findings are, in fact, consistent with offline deliberation studies, which have systematically noted that deliberation is highly context-dependent, with the rules governing the interactions deeply affecting its results (Delli Carpini, Cook and Jacobs 2004; Ganuza 2012).

Researchers have also identified the absence or presence of a moderator as a key factor influencing deliberation (Stromer-Galley and Wichowski 2011; Black et al. 2011; Camaj and Santana 2015). Apparently, when moderators help guide the conversation and ensure that deliberative norms such as mutual respect and rational orientation are enforced, discussions tend to be more deliberative (Albrecht 2006; Stromer-Galley and Wichowski 2011; Black et al. 2011). Even a slight moderation seems to positively influence the deliberative quality of online conversations. For example, Camaj and Santana (2015) compared discussions hosted by the Facebook profiles of US presidential candidates in 2008 and 2012 and found that conversations in moderated sites expressed higher levels of argumentation, even though the moderation of the candidates mostly consisted in agenda-setting and content selection.

In addition, some researchers have claimed the need to research online political discussions taking place in Internet sites that are not explicitly political (Graham 2010; Wright 2011; Black et al. 2011), since most of the scholarship on political deliberation has rather focused on online spaces that are political in nature (such as party websites, government sites, etc.). This is what Wright calls third spaces, that is, online discussion forums “with a primarily non-political focus, but where political talk emerges within conversations” (Wright 2011:254).

Analyzing discussions in third spaces would allow scholars to move towards a more inclusive picture about how the Internet is really affecting political communication and how citizens are actually using it for political dialogue. Some studies have already shown empirically that everyday political talk in online forums whose primary function is not political does have a real democratic value (Coleman and Blumler 2009; Van Zoonen 2007; Graham 2010; 2009).

Camaj and Santana claim that Facebook can be conceived as a third space (2015), since its primary focus is not political. Indeed, Facebook has a primarily non-political orientation, but it also includes spaces where political talk emerges, such as those belonging to political parties, candidates or citizen groups. All of these offer a good starting point for analyzing political talk among ordinary citizens.

A number of studies have analyzed political conversation on social networks, since they constitute “the sociotechnical environments that most closely enable public sphere discourse for those who choose to enter the online ‘salons’ of political candidates” (Robertson, Vatrapub and Medina 2010:29). Indeed, social networks like Facebook allow for direct and open dialogue with established political actors and among citizens themselves. They also allow for activists and supporters to gather around their preferred candidates and parties; and to discuss specific public issues, which is especially relevant for political organizations that lack more traditional offline resources.

The results so far show that conversations on Facebook can be very superficial and are concentrated among a small number of users (Camaj et al. 2009). They very often consist of expressions of non-justified political preferences, and users generally lack any desire to listen, be receptive or enter into dialogue (Valera 2012). Rational argumentation is therefore scarce, since most of the commentators avoid justifying their points of view through reasoned arguments (Sweetser and Lariscy 2008).

Moreover, data reveal that these sites tend to be dominated by like-minded people, generally party activists and candidates’ supporters that share the ideological stances of the political actors hosting the discussions (Fernandes et al. 2010; Valera 2012). For example, (Fernandes et al. 2010) concluded that Facebook was a means for supporters to organize and exhibit their support for their candidate, as well as a place for expressing frustration towards the opponents. Similar results were obtained by Valera (2012), who found a significant homogenous ideological scenario among commentators in the profiles of three Spanish political candidates during a non-electoral period. Indeed, her research revealed that the dissent found in the discussions hosted by the conservative candidate mostly responded to a radical sector of his own party.

But the evidence provided so far about the deliberative quality of Facebook discussions is conflictive. Studies show that discursive freedom is predominantly high, since commentators exhibit civil behavior when entering political dialogue (Camaj et al. 2009; Valera 2012; Camaj and Santana 2015). In fact, comments are generally positive in tone, polite and respectful (Sweetser and Lariscy 2008; Woolley, Limperos and Oliver 2010).

Given the inexistence of absolute criteria for attributing deliberative quality to conversations through such operationalizations, researchers are forced to formulate rather subjective interpretations of their empirical data. In that regard, some scholars conclude that these spaces are hosting political deliberation or that they, at least, can potentially do so (Camaj et al. 2009; Fernandes et al. 2010; Camaj and Santana 2015), while others reject that these conversations meet the deliberative norms (Valera 2012).

Despite these limitations, we still think that deliberative theory offers a pertinent analytical framework for empirically analyzing online political conversations, since it allows researchers to address very relevant concerns about the Internet’s impact on democracy, such as exposure to diversity, discursive freedom, argumentation and reciprocity. Even if we already know that political dialogue on the Internet does not resemble the Habermasian salons where critical public opinion emerged, it is still a useful tool for evaluating phenomena like homophily, endogamy or flaming. And its potential to assess democratic value to online political talk is expanded if we use it in the context of comparative studies.

Moreover, many of the variables employed to operationalize deliberation (such as ideological diversity or discursive freedom) are also useful to provisionally identify alternative democratic functions of these conversations beyond deliberative theory. For example, ideologically homogeneous conversations where the majority of users share political stances (that is, low rates of discursive diversity) won’t allow for deliberation as conceived by deliberative theorists, but they might serve some minority political groups to develop their shared narratives in favorable discursive conditions (intra-ideological discussions). In other words, using deliberative theory as an analytical framework will give us useful insights into which democratic functions can be attributed to online political discussions on Facebook even beyond deliberation.

As a result, we propose a comparative study whose main goal is to assess the democratic value of political discussions hosted on the profiles of three Spanish political candidates running for the General Election of November 20^th 2011. We will operationalize the different conditions of deliberative democracy to conduct our analysis, but we will keep in mind that even if only some of these conditions are met, conversations might still contribute to reinforcing democracy in other ways beyond deliberative standards.

In addition, we propose a comparative approach among conversations hosted on the Facebook walls of three different Spanish political candidates, since recent studies have suggested that online deliberation on Facebook depends on the context in which it occurs and have shown that political ideology does not affect the quality of deliberation (Camaj and Santana 2015). Concretely, we hypothesize that conversation dynamics might vary depending on the size of the political parties represented by the candidates and their uneven centrality in the Spanish political system. That is, we think that the size of the party is an important contextual factor that might affect the Facebook conversation dynamics among followers, as well as their democratic value.

This approach stems from the fact that large hegemonic parties, such as the People’s Party (PP) and the Spanish Socialist Workers’ Party (PSOE) in the Spanish case, tend to receive a great deal of media attention and visibility during the election, as a result of their institutional presence and their centrality in the political system, so that their candidates are more likely to be known by the citizenry. This might facilitate that not only sympathizers and activists but also undecided voters are attracted to these candidates’ Facebook profiles. For their part, minor parties, such as the Union, Progress and Democracy Party (UPyD)2 or the United Left (IU), get much less media coverage and their candidates are not as well known by the citizens. This limited institutional and media presence might prevent non-sympathizers to visit the Facebook profiles of these candidates, so that it is reasonable to expect their online spaces to be more dominated by supporters and activists, since they lack more traditional resources to connect with each other.

The sample of conversations has been selected from four specific days of the campaign: the 8th, 14th, 18th and 21st of November 2011. We needed a sample for the sake of viability, since the profiles of political candidates attract thousands of comments, especially in campaign periods. Thus, we selected all the conversations derived from the posts published during these four days by the candidates Mariano Rajoy (PP), Alfredo Pérez Rubalcaba (PSOE) and Rosa Díez (UPyD).

Initially, the sample consisted of four profiles: the three mentioned above and that of the leader of the IU, Cayo Lara. The four profiles were chosen because they belong to four national parties running in the majority of electoral districts and they constitute a representative sample of the ideological spectrum of the Spanish party system. Moreover, this sample includes two types of political parties according to their size: two large hegemonic political forces who are central in the Spanish political system (PP and PSOE), and two minor parties (UPyD and IU). However, there were no posts published by Cayo Lara for citizens to discuss during the campaign (except for some pictures).

Therefore, our sample focused on all the posts published by the three abovementioned candidates and included a total of 158 posts. However, some of the post had an informative purpose (like publishing detailed information about a campaign event) and did not allow for any debate among citizens. To distinguish between posts that could be debated among users and those that could not, we applied a previous variable called “political density” (Valera 2014b; Sampietro and Valera 2015) to the content of each selected post. This variable allowed us to elucidate if the post contained an ideological or programmatic stance that could be rationally discussed by the citizens.

Only 92 of the 158 initial posts had political density, so that the final sample consisted of 92 posts and 1332 comments, which were unevenly distributed among the candidates’ profiles (Table 1). The total sample of 1332 comments is divided into three subsamples corresponding to Mariano Rajoy (210 comments), Rubalcaba (736 comments) and Rosa Díez (386 comments).

Finally, we limited the number of comments analyzed from each conversation to a minimum of ten and a maximum of fifteen. Given the asynchronous nature of Facebook communication (which allows for potentially infinite conversations), all researchers face the decision of how to limit the sample of comments. We chose to sacrifice the number of comments in every conversation to the maximum of fifteen in order to gather the maximum number of dialogues.

The unit of analysis was the individual message posted by users. Still, these messages were analyzed in the context of a thread. These threads constituted the unit of register in our data base.

In terms of reliability, all the conversations were coded by one researcher. However, the same methodological design is currently being applied to Facebook online political discussions by a four-member team working for a national research project3, and all the variables have been tested. Results show that the percentage of agreement between codifiers is above 90% for discursive freedom and reciprocity and above 80% for the rest of the variables.

When analyzing how often political leaders used Facebook during the campaign, our four-day sample indicates that the socialist candidate showed the largest record of publications, clearly above Rajoy and Díez, who presented more limited publication rates (Table 2). This was probably the result of the socialist’s effort to intensively mobilize the citizenry, which seemed to be the only reasonable way to moderate their expected electoral defeat. Indeed, all the polls predicted an outstanding conservative victory and a socialist defeat during the November 2011 campaign, as well as a significant growth of UPyD. In this unfavorable scenario, the socialist strategy focused on mobilization in order to prevent, or at least moderate, its announced vote loss.

Moreover, Rubalcaba was the only candidate who synchronized his Facebook profile with his Twitter account, which allowed him to launch his messages simultaneously through both social networks during the campaign and be more active online. This synchronization helps explain his intense activity on Facebook. As a result, not every post published by Rubalcaba in his Facebook profile was necessarily conceived of to foster political debate among his followers on this social network. Many posts were designed for Twitter, thus providing users with immediate detailed information about specific campaign events. In this sense, his communicative strategy was more oriented towards informative and propagandistic purposes, allowing his followers to have an updated and constant information flow about the socialist campaign, since half of the posts on Facebook did not show any political density. That is, they did not contain any ideological or programmatic stance that could be rationally discussed by the citizens.

As regards Rajoy and Díez, these candidates presented more limited publication rates (about a fifth and a third of the posts published by the socialist candidate). However, these posts were more frequently oriented towards debate, and allowed for deliberation among their respective followers. Moreover, none of these candidates synchronized their profiles with Twitter. Their messages were therefore specifically conceived of for Facebook.

Hence, Rubalcaba was the most active candidate on Facebook during the campaign thanks to the synchronization with his Twitter account, but only half of his posts allowed for political debate among users; wheras Díez and Rajoy were significantly less active but their posts fostered public discussion more frequently.

It is also worth comparing the number of followers of the three candidates at the beginning of the campaign and after the General Election (Table 3). Unfortunately, we do not have the data at the exact campaign opening (November 4^th). Instead, we present those corresponding to November 8^th, the fourth day of the campaign. These data show that Rubalcaba’s communicative strategy was clearly successful in terms of attracting new followers, inasmuch as his intense Facebook activity during the campaign helped him to almost triple his followers. Also remarkable is the case of Rosa Díez, who was able to gain the support of more than 7000 new users despite being a candidate of a minor party who had much less media visibility. Rajoy, on the other hand, presented the most stable balance of followers before and after the campaign, showing a moderate increase rate of 12%.

The results show that discussions tend to be dominated by like-minded people, since the majority of the comments are favorable to the opinion expressed by the political actors (Figure 1). These findings are consistent with previous studies both in electoral periods (Camaj et al. 2009; Camaj and Santana 2015; Fernandes et al. 2010) and in non-campaign time frames (Valera 2012). This suggests that the profiles of political candidates mostly attract party activists and candidate supporters, thus allowing for direct contact with political actors and expressions of support.

However, there are some differences to be noted. In the case of Mariano Rajoy (PP) and Rubalcaba (PSOE), the two main political parties in Spain, we find some space for dissent in the form of comments expressing opposing views, whereas in the profile of Rosa Díez (a smaller party) there is practically no disagreement to be found among commentators. Her profile also presents the highest level of favorable comments, significantly above the other candidates.

In that sense, the conversations taking place in her profile seem to be much more exposed to homophily than the dialogues hosted in Rajoy’s and Rubalcaba’s profiles, which are slightly more open to ideological diversity. This is probably the result of the fact that both candidates belong to the two largest Spanish political parties, and therefore attract not only activists and sympathizers to their Facebook profiles, but also some undecided voters or disenchanted citizens, who express opposing views to those held by the parties.

In terms of coherence, the findings point out that the majority of comments address the issues introduced by the candidates in their daily posts (Figure 2). These results indicate that most of the participants engage in coherent discussions about ideological and programmatic matters. That is, social networks host online dialogues that are thematically centered on public issues, as proposed by political actors. Moreover, this is the case for all the three candidates analyzed.

However, some users who post comments on the walls of political candidates (especially in the case of Rubalcaba) do so without dealing with the content introduced by the candidates. In those cases, participants either use candidates’ walls to pose specific questions to them (about policy domains, concrete issues affecting users’ directly) or they use them as advertising platforms for content they have created elsewhere (i.e., personal blogs, websites, etc.). That is, incoherent comments mostly serve individual purposes.

Moreover, the largest number of incoherent comments in the case of Rubalcaba’s profile needs to be interpreted in light of his higher publication rate in comparison with the other two candidates. As previously outlined, he posted more intensively than the other two candidates, and many of these posts were merely informative. This might have influenced the commenting dynamics of his followers, who were less systematically invited to collectively debate on specific issues, and therefore showed higher levels of incoherence even when commenting on posts that were conceived to foster discussion.

The comments on Rosa Díez showed the highest level of coherence, close to those on the wall of Rajoy, and clearly above those on Rubalcaba’s profile, which deviate much more often from the issues proposed. In other words, conversations in Rosa Díez’s profile (a candidate of a minor party) show the highest level of both ideological and thematic cohesion, while the dialogues on Rajoy and Rubalcaba’s profile (candidates of major parties) present slightly lower levels of both variables. Taken together, these results suggest that the conversations on Díez’s wall were more often held among party activists, who are more willing to debate on public issues, compared to those hosted on the walls of candidates belonging to larger parties, where thematic and ideological cohesion is not so intense.

The findings show that an overwhelming (and quite evenly distributed) majority of participants do not provide any reasoning for the claims they make (Figure 3). Indeed, quite the opposite occurs: most users join the dialogues to express individual preferences, but have no intention of rationally justifying their specific points of view.

Taken together with the results on ideological diversity, these findings suggest that most of the users participate in the candidates’ walls to convey their ideological support in a rather vague way (Sweetser and Lariscy 2010). In other words, these third spaces mostly attract activists’ non-reasoned support messages during the campaign, as suggested by previous research (Camaj et al. 2009; Camaj and Santana 2015).

Only a very small number of users engages in rational discussion and provides reasoned claims and, more remarkably, practically none delivers evidence to support their claims. In that sense, both the presence of argumentation and the absence of evidence are quite evenly distributed among candidates’ profiles. However, the profile with the highest proportion of reasoned comments is that of Rosa Díez, whose participants engage more often in rational argumentation than those commenting on the walls of the other two candidates. Indeed, the results indicate that political talk on Díez’s profile is more ideologically homogeneous, coherent and rational-oriented.

The findings on reciprocity clearly show that these online forums, despite being hosted in the profiles of political candidates, are mostly dominated by citizens, since the candidates very rarely take part in them (Figure 4). The data demonstrate that Rajoy (PP) and Rubalcaba (PSOE), in particular, do not engage in political talk with citizens. Instead, their activity consists of content moderation and agenda setting (Camaj and Santana 2015). However, the UPyD candidate, Rosa Díez, does occasionally respond to Facebook users, though in a very small proportion. Reciprocity among candidates and citizens is therefore very infrequent.

Regarding soft reciprocity among the participants themselves, the data show that at least a minority of users formally speaks to each other. That is, users interact with each other more than they do with candidates, especially those posting on the socialist profile. It should be noted, however, that the discussions on Rajoy’s and Díez’s profile score more poorly in terms of users’ responses.

However, if we take a closer look at these interactions, it becomes clear that only a minority of them incorporate arguments provided by other participants in previous comments. Argument incorporation is less frequent than users responding to each other formally, except for the case of Rosa Díez’s wall, where, surprisingly, users tend to incorporate arguments more often than to respond to each other. Again, this is probably a result of the fact that Díez’s conversations are more dominated by party activists who share ideological stances, so that it is easier to find argument incorporation in intra-ideological discussions where no fundamental disagreement is to be found.

In sum, the vast majority of commentators in all three analyzed profiles do not show any intention to exchange points of view or arguments. These findings are consistent with previous studies that have shown the lack of responsiveness and listening of online political discussions (Wilhelm 1998; Dumoulin 2002; Jensen 2003; Valera 2012). Interestingly enough, monologue is much more frequent than dialogue in online political talk (Freelon 2010). All of this suggests that users mostly take part in these online spaces to make their voices heard by both political candidates and other users (but not to take part in rational argumentation).

Finally, the results on discursive freedom demonstrate that most commentators use a neutral and respectful language (Figure 5). Moreover, discursive freedom is quite evenly distributed among profiles, since harsh language is marginal in all of them.

Still, some slight differences should be noted. The discussions on Rosa Díez’s profile score higher in discursive freedom, since there is almost no room for harsh language, insults or flaming in these conversations. These results are probably associated with higher ideological homophily, since like-minded people are less likely to engage in flaming and uncivil behavior.

Rajoy’s and Rubalcaba’s walls, however, include some small proportion of uncivil behavior in their conversations, so that discursive freedom in these spaces is more restricted. This is probably related to the presence of disagreement, which leads more easily to cross-cutting dialogues and confrontation.