Interactivity and evaluation of public figures in social media

Interactivity in websites refers to the intrinsic features of the medium or technology that are different - conceptually, at least - from users’ perceptions of interactivity (^{Sundar, 2007}). Then, this perspective compares the medium specific features such as the availability of links, scrolling bars, instant messaging, etc., to the perceptions that users hold about those features. Although two-way communication on the Internet has been available since the beginning of the 90s, favoring communication between people or between individuals and interfaces, users’ perceptions about interactivity levels may vary according to interpersonal factors (e.g., personal preference for face-to-face interactions; perceived usefulness of the interfase), or by the influence of other users (e.g., chosing Whatsapp instead of Skype). In any case, interactivity has been associated to positive results such as a better understanding of message content and changes in attitudes towards health communication (^{Kim & Stout, 2010}) or marketing websites (^{Fiore, Kim & Lee, 2005}).

Research in social media has also confirmed a relationship between interactivity and positive evaluations, but attention has traditionally been paid to the users’ subjective perceptions about the account owner instead of the interactivity with the website features (^{Rafaeli & Sudweeks, 1997}). Accordingly, emphasis has been placed on the impressions that users gain from observing the interactions between the public figure and his/her followers (^{Kruikemeier, 2014}; ^{Walther et al., 2008}). Political interactivity in social media has been regarded as the replies made by candidates or political figures to messages published by citizens on their personal social network pages (^{Evans, Cordova & Sipole, 2014}).

The scarce literature on interactivity in Twitter has reported several mechanism models explaining the way in which politicians are successful communicating with their potential voters, mobilizing them to vote for them or to participate in political activities either in online or offline contexts. In this sense, interactivity is conceptualized as “the extent to which a politician’s Twitter communication represents two-way conversations, as opposed to one-sided public addresses” (^{Lee & Shin, 2012, p. 515}). Users develop positive impressions about politicians who reply to the messages posted by their online audiences as opposed to those who do not get involved in this type of conversations (^{Utz, 2009}).

These impressions take place on the basis of social presence perceptions produced after watching interactions. Social presence (^{Short, Williams & Christie, 1976}) represents the extent to which a person is “salient” in the communication exchange, and is related to feelings of closeness, intimacy, and proximity. Although the content of the communication exchange could be erratic or negative (e.g., insults between candidates and users), the perceptions of interactivity foster the consciousness that the public figure actually responds to the users’ comments, thereby increasing the feeling that he/she is socially present (^{Nowak & Biocca, 2003}).

In their study, ^{Lee and Shin (2012)} introduced messages posted on a politician’s mock-up Twitter page in which he replied to followers about specific aspects of his personal life, such as his favorite sporting activities, places he preferred to spend during the holidays and other information (e.g., @politician “Actually, my fourth-grade son and I are diehard soccer fans”; RT@user2 “What’s your favorite sport?”). As hypothesized, the participants exposed to that page evaluated the politician’s interactivity with users more positively than the same page in which he had little interactivity. Therefore, in line with the insights provided by ^{Lee and Oh (2012)}, the present study conceptualizes interactivity as the user’s perception about the two-way communication process that a politician holds with other users in contrast to messages addressing the general public. Moreover, this study assumes that the positive effect of interactivity is related to its frequency: the more replies he or she gives, the better the political figure will be evaluated.

Identity and party identification

One of the most significant influencing factors on electoral decisions is party identity, which, from a social psychology perspective, is seen as an affective predisposition of closeness towards a political party based on feelings of affinity or proximity towards social groups associated to that party (^{Campbell, Converse, Miller & Stokes, 1960}); and which emerges as a result of pre-adult socialization processes and assimilation of new events (^{Green, Palmquist & Schickler, 2002}). Identification levels with the political party grow from these feelings of affinity: for example, in American politics, the closer a potential voter feels towards groups associated to the Republican Party (e.g., businessmen, White evangelical Protestants) the greater his/her level of identification with that party and its members will be (^{Goren, 2005}).

Alternative sociological perspectives suggest that party identification emerges from belonging to certain social classes and conditions (e.g., religion, socioeconomic status, place of residence). The divisions between these classes are thought to inform the ideologies and pragmatic principles of each party (^{Fleury & Lewis-Beck, 1993}). Another perspective suggests that party identities are not impermeable to changes: people tend to adjust their loyalties and to analyze their political affiliation on a reasoned basis of evidence and experience. In other words, they change their identities from retrospective and prospective assessments of the government’s performance, the state of the economy, the extent of representativeness, and others (^{Fiorina, 1981}). As a matter of fact, it has been suggested that the most evident features of party identification during the Mexican democratic transition in 2000 were: 1) the identity shifts that occurred as a product of evaluations of the government’s performance, and 2) the widely-spread “anti-PRI” feeling originated by the long permanence of this party in power (^{Estrada, 2005}; ^{Guardado-Rodríguez, 2009}).

Although these perspectives enrich the notion of party identity, the argument of identity change or “adjustment” (often mirrored by tendencies in longitudinal surveys) tends to weaken when the measurement error is subtracted from scores, suggesting that party identities are more resistant to change than expected (^{Goren, 2005}). On the other hand, the sociological argument for explaining party identification as a product of social divisions or contextual cleavages (e.g., left-right wing ideologies, liberal-conservative principles, and others) makes sense in consolidated democracies but not in relatively younger ones such as in Latin American countries, in which party identification is probably more a product of a “transactional” response or a clientelist practice than a real conviction about the party’s programmatic agenda (^{Morales-Quiroga, 2016}).

The present work aims to return to the social psychological notion of party identification and to focus on its affective aspect. Consequently, this study defines party identification as a positive attitude of closeness towards a specific political party and its members. This attitude may vary according to individual experiences of closeness or affiliation, more than a reasoned ideological conviction; and it is independent of formal, official or institutional membership (^{Campbell et al., 1960}). This notion of identity mirrors the concept of identity in ^{Lee and Oh (2012)}, and other instruments applied in the Mexican context such as the CIDE-CSES political survey.

Social identity and party identification in Twitter

One of the most solid models explaining the formation of a group identity is ^{Tajfel and Turner’s (1986)} social identity theory (SIT). This approach considers that people define themselves in terms of their personal traits or their membership of social groups or categories. A group is conceptually seen as a set of individuals that perceive themselves as members of the same social category and share an emotional attachment to their group and memberships. Thus, from the “minimal group” paradigm, the authors assert that the mere awareness of group divisions is enough to make people take part in one of those groups, thus assuming its norms and seeking a positive social identity. This is to say, members will tend to seek social prestige for their own group and disregard other groups. Based on this principle, they will tend to favor “in-group” members and to disdain “out-group” members.

It is important to highlight that in-group members are favored as far as they are seen as prototypical of their group, that is to say, the extent to which they represent in-group traits and norms accurately. For instance, a prototypical member of a right-wing party is the person who embodies the interests of businessmen, sticks to the practices and principles of the Christian religion, and express his/her disagreement with the intervention of the government in market regulation. Nevertheless, the extent to which a member is seen to be prototypical of the group is essentially subjective, and the evaluation tends to be more rigid when it is made by in-group members with a high level of group identification (^{Turner, Hogg, Oakes, Reicher & Wetherell, 1987}). Hence, party identity will be analyzed from this perspective: those with a high level of party identification will tend to evaluate other members’ prototypicality in a stricter way than those who have lower levels of party identification.

The research on computer-mediated communication suggests that the aforementioned process also takes place on the Internet. The social identity model of deindividuation effects (^{Postmes, Spears & Lea, 2002}) postulates that introducing a salient social identity (e.g., telling users the they take part in a virtual group) among users that remain anonymous, is sufficient to divert their attention from their personal traits and to activate awareness of group membership. Similar to ^{Tajfel and Turner’s (1986)} original model, when a social identity is relevant to users they tend to categorize themselves as in-group members, adopting its norms, favoring other members, and disregarding out-group members (^{Lea, Spears & de Groot, 2001}).

Similarly, there is evidence of the moderating role of social identity when users come to assess people perceived to belong to the in-group or the out-group in social media. In ^{Mou, Miller and Fu (2015)} the participants sharing membership in the category “professors” evaluated the social network profile of an in-group member as more credible and attractive than the evaluations made by out-groupers (students). Moreover, ^{Lee and Oh (2012)} pointed out that the publication of personalized messages by a politician in Twitter (e.g., information related to their personal aspects) provoked less favorable impressions about him among in-group members with higher party identification, as they consider that the candidate seemed to be following a personal agenda of self-promotion rather than a political agenda linked to the party ideology. Conversely, users with lower levels of identification evaluated the personalized messages more positively, and consequently, showed greater intentions to vote for the candidate. In summary, the interactivity of public figures in social media is related to positive evaluations of them made by users. Therefore, for this study it is expected that Twitter users perceive a politician’s interactivity as a positive trait: more interactivity will lead to better assessments, and greater intentions to vote for him/her. Nevertheless, these impressions will be moderated by the users’ party identity and the extent to which ‘they identify with their party.

An additional aspect of importance is the extent to which a candidate is relevant for users. Because this study took place during a political campaign, the presence of candidates in traditional and digital media was ubiquitous, and their relevance for citizens increased as the election day drew closer. This relevance may interfere with the model introduced given that the effect of interactivity might overlap with party identity and identification levels. In other words, how can we be sure if the positive effect is produced by the candidate’s interactivity or by his/her personality and/or party identity salience? In order to address this potential overlap, the present study will include a control condition in which the candidate shares membership with the same party as users’ but whose campaign runs in a different State and does not hold any interactivity with her followers.

Accordingly, this study’s predictions are formulated as follows:

A candidate’s interactivity in Twitter will affect the way she is evaluated: H1) a greater degree of interactivity will produce, overall, more positive evaluations about the candidate than a lower level of interactivity and than evaluations of the candidate in the control group.

The party identity of users will affect the way in which a candidate in Twitter is evaluated: H2) participants with the same party identity as the candidate (in-group) will evaluate the candidate more positively than participants who do not share the same party identity (out-group).

Users’ party identification levels will affect the way in which the candidate’s interactivity in Twitter is evaluated: H3) those with a high degree of party identification will evaluate the candidate’s high levels of interactivity less positively than those with a low level of party identification, whereas both interactivity conditions will be evaluated more positively than the control group.

Participants

139 university students from different disciplines and backgrounds participated in the study (53 male, 86 female; age M= 20.14, SD= 1.96).

Procedure

Volunteers participated individually in a computer room in the months before the end of the 2015 Nuevo Leon election campaign (February to May). The campaign initially included 10 contenders and concluded with only 7 contenders. Following procedures in previous studies, participants were asked whether: a) they had a Twitter account and, b) if they knew the candidates. As most students answered affirmatively to both questions, they were led to continue with the rest of the session. First, students filled out a questionnaire with demographic information, party identity, and attitudes toward politicians and questions regarding time spent in Twitter. Then, participants were presented with the Twitter account of the candidate representing the PRI² party and were asked to read all the messages posted on her profile page (10 messages) before filling out a second questionnaire with additional scales. The Twitter mock-up page was, in fact, the candidate’s real account but the messages were re-organized for this study’s purposes (replies).

Design

A 3 (interactivity: high, low, control) x 2 (party identity: in-group, out-group), x 2 (identification level: high, low) experimental design was introduced in order to check their effects on evaluations about the candidate. In the high-interactivity condition the candidate replied individually to six messages posted by supporters, whereas the low-interactivity consisted of two replies only; the remaining messages posted comprised status updates describing her political agenda. An example of a reply to a particular message is described below:

@User: Our commitment is also with you [candidate], working hand in hand for Nuevo Leon

@Candidate: I appreciate your trust in me a lot my friend!

In contrast, an example of a non-interactive message or post addressing the general public is shown below:

@Candidate: As a governor, I will boost social programs for crime prevention in the district

The control group condition (less relevant candidate) included the Twitter account of the PRI candidate for the Sonora State as a stimulus. Similar to the other conditions, 10 messages were included describing only status updates (e.g., @Candidate: Enjoying the streets and the taste of Banamichi, Sonora). This condition was introduced in order to contrast interactivity effects and party identity.

Measures

Party identity was measured from participants’ self-reports: they were presented with the political parties of the 10 contenders (including an independent candidacy) and were asked to select the party they felt more identified with. The in-group was conformed by those who chose the same party of the candidate (PRI) whereas the out-group comprised all participants that chose another candidate. The level of identification was measured with a 10-point scale and recoded taking the mean as cut-off point: scores above and below represented users with higher or lower party identification. Although the number of participants comprising the in-group was notably smaller (10.1%) than participants comprising the out-group (89%), no significant differences were observed among the experimental conditions in terms of party identity (Χ ²[2] = 0.40, p= n.s.), or identification levels (t[137] = 0.24, p= n.s.). In addition, previous attitudes toward candidates were controlled with a 10-point scale going from 1 (“I dislike her very much”) to 10 (“I like her very much”).

Overall evaluation of the candidate was measured with items in a 7-point semantic differential as in ^{Lee and Shin (2012)} (e.g., intelligent-unintelligent, honest-dishonest, likable-unlikable). A factor analysis test showed two dimensions that were labeled as “political aptitudes” and “personal aptitudes”. Finally, voting intention was measured with two items: “I would vote for this candidate if the elections took place tomorrow” and “I would never vote for this candidate” (r= .79).

Results

In order to test the hypotheses formulated, an interactivity x party identity x party identification levels multivariate analysis of variance (MANOVA) was performed on the overall evaluation of the candidate and intentions to vote for her, also controlling for previous attitudes, participants’ gender, and time spent on Twitter. The first hypothesis predicted better overall evaluations of the candidate with higher levels of interactivity: although no multivariate effects were observed for interactivity, a significant one-way effect of interactivity on the candidate’s personal aptitudes was detected (F[2, 104] = 4.14, p = .04). The participants evaluated the candidate with more interactions as having better personal attributes (e.g., likable, attractive; M = 4.50, SD = 1.16) than the same candidate with lower levesl of interactivity (M = 4.14, SD = 1.10). Nevertheless, the candidate in the control group (i.e., different State elections) was evaluated slightly better than the more interactive candidate (M = 4.84, DT = 1.29). Therefore, H1 was corroborated only partially (Figure 1).

Source: The author.

Figure 1 Evaluation of the candidate’s personal aptitudes from her interactivity in Twitter 

The second hypothesis predicted better the evaluations by participants sharing the same party identity as of the candidate. A significant multivariate effect was observed for the participants’ party identity on the candidate’s political attributes: Wilk’s λ = 0.92, F(3, 103) = 2.90, p = .03. Specifically, scores in the in-group (M = 4.94, SD = 1.14) compared to the scores in the out-group (M = 3.67, SD = 1.27), showed that in-group participants perceived their candidate as having better political attributes (F[1, 105] = 4.63, p = .03, η_p ² = 0.04) (Figure 2).

Source: The author.

Figure 2 Evaluation of the candidate’s political aptitudes from sharing membership with the participants’ party identity 

Moreover, in-group members showed more intentions to vote for her (M = 5.00, SD = 1.32) than out-group members (M = 2.29, SD = 1.60) (F[1, 105] = 6.30, p < .001, η_p ² = 0.05). Therefore, H2 was corroborated.

Source: The author.

Figure 3 Intentions to vote for the candidate as a result of sharing membership with the participants’ party identity 

The third hypothesis predicted less positive assessments of the candidate among in-group participants with higher identification levels compared to in-group participants with lower identification levels. No significant multivariate effects were observed from the interaction between party identity and identification levels on the overall evaluations. Instead, a one-way effect on the participants’ willingness to vote for the candidate was detected (F[5, 105] = 2.38, p = .04, η_p ² = 0.10). Contrary to the predictions, those with lower levels of identification showed a greater intention to vote for the candidate with less interactivity (M = 5.83, SD = 1.83) than the same candidate with more interactivity in Twitter (M = 4.04, SD = 2.24), and also than the candidate in the control group (M = 3.75, SD = 1.83) (Figure 4).

Source: The author.

Figure 4 Intentions to vote for the candidate as a result of sharing membership with the participants’ party identity: low level of identification 

Conversely, those with a high level of identifcation showed greater intentions to vote for the candidate with higher levels of interactivity (M = 6.09, SD = 1.59) than for the less interactive candidate (M = 2.04, SD = 1.75). Surprisingly, and contrary to the expected outcome, intentions to vote were higher in the control group than in the high interactivity condition (M = 6.40, SD = 1.75) (Figure 5). Hence, H3 was not corroborated.

Source: The author.

Figure 5 Intentions to vote the candidate as a result of sharing membership with the participants’ party identity: high level of identification