Political discourse on social networking sites

Galley Proof

10/05/2013; 9:21

File: IP303.tex; BOKCTP/xhs p. 1

Information Polity 00 (2013) 1–20 DOI 10.3233/IP-130303 IOS Press

1

2

3

1

Political discourse on social networking sites: Sentiment, in-group/out-group orientation and rationality1

4

Scott P. Robertson∗ , Sara Douglas, Misa Maruyama and Bryan Semaan

5

Information and Computer Sciences Department, University of Hawaii, Honolulu, HI, USA

6

14

Abstract. The news feeds of two U.S. politicians’ Facebook sites were examined across 22 months leading up to an election in order to explore changes in social-network-mediated public political discourse over time. Changes over time were observed in who was being addressed and in the affective valence of comments. A complex flow of attention between in-group and out-group concerns was observed with in-group comments dominant both in early and late phases. Also, positive comments decreased and negative comments increased over time. These phenomena, dubbed “reflection-to-selection” and “converging sentiment”, were refined to explain the observed nonlinearities. The flow of rational versus affective comments in politicians’ Facebook data across time was also explored. Comments reflecting cognition were more prevalent at all times than comments reflecting affect, but their distribution also varied in complex ways over time. Finally, the concept of “potential public sphere” in contrast to “realized public sphere” in virtual spaces is introduced.

15

Keywords: Digital government, social networking, e-participation, e-citizenship, virtual public sphere

16

1. Introduction

17

1.1. The internet and political activity

18

Since social networking sites (SNSs) first appeared on the Internet, their use has consistently increased and the purposes for which they are used have broadened. A recent survey conducted by the Pew Internet and American Life project [37] reports that 65% of all Americans say they use SNSs, up from 2% since 2005. A 2011 Nielsen study [47] reports that Americans spend 23% of their Internet time using blogs and social networks, a larger percentage than any other online activity. Americans spend more time on Facebook than any other site, and Facebook is the most popular application across all ICT platforms [21]. An increasingly important use of SNSs is for participation in civic and political activities. Hampton et al. [21] report that, controlling for demographic characteristics such as education and income, Internet users are much more likely than non-Internet users to have been involved in civic/political activities such as attending a political rally (2.39 times more likely), trying to influence a vote (75% more likely), and

7 8 9 10 11 12 13

19 20 21 22 23 24 25 26 27

1

This work is based on an earlier work: Political dialog evolution in a social network, in Proceedings of the 13th Annual c International Conference on Digital Government Research (dg.o ’12), ACM, 2012. http://doi.acm.org/10.1145/2307729. 2307737. ∗ Corresponding author: Scott P. Robertson, Information and Computer Sciences Department, University of Hawaii, Honolulu, HI 96815, USA. E-mail: [email protected]. c 2013 – IOS Press and the authors. All rights reserved 1570-1255/13/$27.50

Galley Proof

2

10/05/2013; 9:21


S.P. Robertson et al. / Political discourse on social networking sites

33

actually voting or reporting intention to vote (53% more likely). There has been considerable investigation and controversy about whether Internet use generally, and SNS use specifically, has an impact on civic participation [1,12,48,49], but data shows that SNS use and civic activity are certainly related since frequent Facebook users, when compared to other Internet users, are 2.5 times more likely to attend rallies, 57% more likely to have tried to influence a vote, and 43% more likely to have reported voting or intending to vote [21].

34

1.2. Goals of this article

35

41

In this article we explore comments posted on the Facebook sites of politicians running for office. This is an area where civic discourse should be prevalent. Our goal is to describe the types of discourse that occur in the virtual public sphere of SNSs and show how this discourse changes over time. We reexamine two previous phenomena that we have observed and reported elsewhere [59,63], reflection-toselection and converging sentiment, to find more subtle changes over time. Further, we examine rational and affective components of the dialogue taking place on politicians’ official Facebook pages. We ask whether both take place, in what proportions, and whether there are changes over time.

42

2. Literature review

43

2.1. ICSs and the public sphere

44

52

The social networking components of many online sites are often described as a new type of sociotechnical public sphere such as Habermas [7,20] described in non-Internet contexts. Researchers have asked whether true civic discourse takes place online [12,14,16,17] and whether the notion of a public sphere is changed by socio-technical mediation [2,4,6,8,9,75]. Recent empirical work has sought to document as public spheres many types of online spaces such as political blogs [10,15,70,73] and microblogs [41], discussion forums [52], newsgroups [24], online videos [44,45], and SNSs [32,40,42,61, 62,78,79]. The ideal of a public sphere for political discourse is that it will support rational discussion in which multiple viewpoints are heard and consensus is reached, although such ideal discussions may rarely occur in real or virtual contexts.

53

2.2. Why people use SNSs

54

When we examine discourse over time, we must ask what brings people back or what sustains their conversation. Joinson [26] provides an early study of the reasons that people give for using Facebook. He found seven prominent motivations:

28 29 30 31 32

36 37 38 39 40

45 46 47 48 49 50 51

55 56 57 58 59 60 61 62 63

– – – – – – –

Keeping in touch, Interacting with others who have shared interests and identities, Viewing photographs, Using applications like games, Finding new friends and acquaintances through social investigation, Browsing friends of friends, called “social network surfing”, Informing others via status updating.

Galley Proof

10/05/2013; 9:21



3

90

The second motivation is related to joining groups, organizing activities, and meeting like-minded people and tends to be associated with greater time spent in SNSs. Lampe et al. [32] referred to this activity as “social browsing”. Interestingly, learning or discovering something outside of the arena of information about friends is not included among these motivations, except to the degree that it is found in the shared interests and identities motivation. Lampe et al. [33] showed that the motivations of users changed during their involvement with groups. They emphasized that the motivations for joining a group differ from the motivations for remaining in a group. Often the change is from information seeking in the beginning to issues such as maintaining connectivity and gaining satisfaction and a sense of belonging in later stages. Dholokia et al. [13] distinguish among several motivations for participating in online communiities: purposive value, self discovery, maintaining interpersonal connectivity, gaining social enhancement, and deriving entertainment. These vary across different types of communities and through time. Ren et al. [57] distinguish between identity-based reasons for participating on online communities and bond-based reasons [53,54,71,72]. Common identity underlies users’ participation in communities when they feel that they have shared interests. Members of an identity-based group will not necessarily expand their relationships with others beyond the shared concern of the group. Common bond underlies users’ participation in communities when they are already bonded in some way such as friendship or family. Members of a bond-based group usually share common interests, but their participation in the group is not limited to the shared interest(s). Ren et al. [57] showed that the design of the sociotechnical environment can differentially influence its usefulness for identity-based versus bond-based communities. The degree to which social browsing operates to expose SNS users to new ideas will depend on the context in which the browsing takes place. Whether or not political discourse is broadened by exposure to diverse ideas in computer-mediated settings has been a subject of debate [5,18,25,67–69,74]. On the one hand, online discourse may broaden exposure to a diverse set of ideas [55,76,77,79]. On the other hand, it has also been observed to narrow exposure to ideas in an echo-chamber of like-minded friends (i.e. “homophily”) [1,32,43,46,48,49].

91

2.3. Rationality and emotion in political discourse

92

Political decision making has both rational and emotional components. A rational-actor model of political decision making emphasizes the gathering of information, weighing of choices, and reasoning about costs and benefits of various decisions. While such models have been prevalent in political science for many decades [58], they have been challenged by theorists and researchers coming from psychology and other behavioral social sciences [34,65,66]. One challenge involves the pragmatics of making complex decisions that require significant cognitive resources. Lau [34] points out that many people are unmotivated to put in the cognitive effort required to make complex decisions that can be made effectively enough by using short-cut heuristics. This view falls into a broader category of bounded rationality theories [23,27,64–66] in which actors weigh the complexity of a problem against their cognitive and other resources. A popular theory in political psychology is the “on-line running tally” idea that people create an initial, often schema-based concept of a candidate which they then use as a comparator when new information is encountered [35,36,56]. The candidate schema is updated as necessary while the information used for updating might be forgotten. Social psychologists have spent considerable effort studying the role of emotion in the political arena [19,38]. Marcus et al. [39] propose that political judgment is a balance between habit and reason that is mediated largely by emotion. Studies of the impact of media on decision making have also

64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89

93 94 95 96 97 98 99 100 101 102 103 104 105 106 107

Galley Proof

4

10/05/2013; 9:21



Table 1 Reflection-to-selection: Observed correlations with time of number of first-person and third-person pronouns for political candidates in two studies (2008 data df = 46, 2010 data df = 20)

First-person (I, we) Third-person (s/he, they)

2008 U.S. presidential election Obama Clinton McCain −0.87* −0.52* −0.78* +0.84* +0.80* +0.88*

2010 California gubernatorial election Brown Whitman −0.27 −0.43* +0.42* +0.16

Table 2 Converging sentiment: Observed correlations with time of number of positive and negative comments for political candidates in two studies (2008 data df = 46, 2010 data df = 20)

Positive comments Negative comments Positive − negative

2008 U.S. presidential election Obama Clinton McCain −0.86* −0.64* −0.82* +0.90* +0.52* +0.75* −0.90* −0.64* −0.82*

2010 California gubernatorial election Brown Whitman −0.20 −0.48* +0.17 +0.61* −0.21 −0.61*

110

emphasized the role of impression formation and manipulation in forming political judgments [80]. In this paper we begin an examination of rational and affective components of discourse in SNSs and ask how these two types of discourse change over time.

111

2.4. “Reflection-to-selection” and “converging sentiment”

112

Robertson and his colleagues [59–63,73] have examined whether the comment areas of politicians’ Facebook pages serve as socio-technical public spheres that encourage exploration of diverse ideas and mutual exchange of information [10,11]. A politicians’ comment area on a SNS supports an identitybased group. However, participants in the comment area are both in-group members (supporters of the candidate) and out-group members (opponents of the candidate). The overarching shared interest is an upcoming election. Within that broad interest group are opposing subordinate interest groups for each candidate. In previous work [59,63], Robertson and his colleagues found evidence for several general types of changes in discourse on politicians’ Facebook sites over the months leading up to an election. These changes were related to group identity and to the valence of conversation. In virtually all cases, the use of first-person pronouns (“I” and “we”) and second person pronouns (“you”) decreased over time while the use of third-person pronouns (“he”, “she”, and “they”) increased over time. Table 1 shows the observed correlations of number of pronouns in these categories with time from the two studies. Further examination of the pronoun “you” showed that the decrease over time was for instances where the candidate was being directly addressed and not when “you” referred to other people. This pattern was interpreted as evidence for a “Reflection-to-Selection” model in which the community was inward facing and engaged in defining a shared identity during the early stages of a political campaign (the “Reflection” stage), but then later turned outward toward others and the opponent (the “Selection” stage). A sentiment analysis of the affective nature of the comments showed that dialog tended to be more positive than negative, but that positive comments decreased over time while negative comments increased over time such that they were almost equal in proportion when the elections occurred. Table 2 shows the observed correlations of positive and negative affect terms with time from two studies. This pattern was interpreted as evidence for a “Converging Sentiment” phenomenon in which early identity definition was largely a positive activity focused on shared values and ideas but, once formed on both sides, turned into an identity contrast activity that was more negative.

108 109

113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136

Galley Proof

10/05/2013; 9:21



5

160000 140000 120000 100000 Meg Whitman

80000

Jerry Brown

60000 40000 20000 0 1

3

5

7

9

11 13 15 17 19 21

Fig. 1. Mean number of words in Meg Whitman’s and Jerry Brown’s Facebook news feeds in the months leading up to the 2010 California Gubernatorial election.

140

In this article we re-examine the data collected from the 2010 gubernatorial election in California [63]. Specifically, we looked for non-linear aspects of the pronoun and sentiment data. We also added observations about rational and non-rational discourse as reflected by words describing cognitive mechanisms and affect.

141

3. Method and data set

142

User comments from the Facebook news feeds of the two candidates in the 2010 California gubernatorial race, Jerry Brown and Meg Whitman, were examined for this study. Neither candidate was an incumbent although both were well known in California. Brown was serving as the State Attorney General of California at the time of the election and had been the Governor from 1975–1983. Whitman was a former CEO of eBay. Public comments open to any Facebook user or authorized application were collected from the February 9, 2009 to November 13, 2010 time period. The election itself took place on November 2, 2010. Comments were extracted using Facebook’s Open Graph API in compliance with Facebook rules and Open Graph restrictions. There were 9,442 comments from 6,804 individuals collected from Jerry Brown’s Facebook site and 11,787 comments from 5,039 individuals collected from Meg Whitman’s Facebook site. Figure 1 shows the number of words appearing per month on the two candidates’ Facebook sites in the months leading up to the election. Linguistic Inquiry and Word Count (LIWC) [50,51] was used to analyze the corpus. LIWC provides basic word counts and percentages of words in several part-of-speech and semantic categories. Semantic categories were derived empirically by the LIWC developers, and word categories have been psychometrically tested [51] and used previously in a variety of analyses of discourse [3,22,28–31]. The reflection-to-selection phenomenon was examined by using the person of pronouns (first-person, second-person, and third-person). The converging sentiment phenomenon was examined by finding the valence (positive or negative) of words reflecting affect. The rationality question was addressed by examining words in the cognitive mechanism category and words in the general affect category. Cognitive mechanism words are terms that reflect discussion of mental processes, insight, and analysis (e.g.

137 138 139

143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162

Galley Proof

6

10/05/2013; 9:21



193

“think”, “know”, “cause”). Affect words are terms that express emotion and feeling (e.g. “feel”, “like”, “hate”). The ordinal sequence of months leading up to the election was coded with the numbers 1–22 where 1 is the farthest month preceding the election and 22 is the month of the election. Multiple regression analyses were performed using the linear time sequence as one predictor and its quadratic transformation as a second predictor. The quadratic trend in regression tests for a line which curves in one direction. Separate analyses were performed on the following criterion variables: – Whitman’s use of first-person pronouns. – Brown’s use of first-person pronouns. – Whitman’s use of third-person pronouns. – Brown’s use of third-person pronouns. – Whitman’s use of words reflecting cognitive mechanism. – Brown’s use of words reflecting cognitive mechanism. – Whitman’s use of use of words reflecting affect. – Brown’s use of words reflecting affect. – Whitman’s use of words reflecting positive affect. – Brown’s use of words reflecting positive affect. – Whitman’s use of words reflecting negative affect. – Brown’s use of words reflecting negative affect. For the content analysis, a random sample of 100 posts was taken for each candidate from four different time periods. The time periods were determined by a combination of number of posts within the period (in order to guarantee a large enough sample pool in each period since the number of posts greatly increased in the months just before the election) and important events. The sample periods were February 9, 2009–December 31, 2009 (691 posts for Brown and 465 posts for Whitman covering the beginning of the Facebook activity through the end of the year), January 1, 2010–June 8, 2010 (2122 posts for Brown and 1901 posts for Whitman leading up to the primary election), June 9, 2010–September 28, 2010 (2667 posts for Brown and 3714 posts for Whitman), and September 29, 2010–November 13, 2010 (3971 posts for Brown and 5707 posts for Whitman immediately before and after the election). This resulted in a sample of 800 posts, 400 for each candidate, which were then categorized independently by two of the authors (SD and MM). Agreement between the experimenters was above 80% for each category. Disagreements were reconciled for all sampled posts.

194

4. Results

195

4.1. Reflection-to-selection

196

Table 3 shows the mean number of first-person and third-person pronouns used by both candidates combined across the 22 time periods in this study. A 2 × 2 repeated-measures ANOVA was performed using time period as if it were a subject factor. There was significantly more use of first-person pronouns (mean = 4.11 percent) than third-person pronouns (mean = 1.73 percent), F (1,21) = 121.02, p < 0.0001, and significantly more pronoun use by Whitman in contrast to Brown, F (1,21) = 4.84, p < 0.039. However, there was also an interaction between the pronoun variable and the candidate F (1,21) = 5.48, p < 0.029. Examination of the means suggests that although the percentage of third-person pronouns was about the same for both candidates (means = 1.68 and 1.78 percent for Whitman and

163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192

197 198 199 200 201 202 203

Galley Proof

10/05/2013; 9:21



7

Table 3 Mean percentage of first-person and third-person pronouns in two political candidates’ Facebook news feeds combined across 22 months preceding an election

Table 4 Mean percentage of cognition and affect words in two political candidates’ Facebook news feeds combined across 22 months preceding an election

2010 California gubernatorial candidates Pronoun person Meg Jerry Mean category Whitman Brown First-person 4.40 3.83 4.11 Third-person 1.70 1.78 1.73 Mean 3.04 2.80

2010 California gubernatorial candidates Cognition/affect Meg Jerry Mean category Whitman Brown Cognitive mechanism 15.95 14.21 15.08 Affect 5.66 5.66 5.66 Mean 10.80 9.93

222

Brown respectively), the percentage of first-person pronouns was higher for Whitman (mean = 4.40 percent) than for Brown (mean = 3.83 percent). Figure 2 shows scatterplots of the percentage of first-person pronouns and third-person pronouns in the comments in both candidates’ Facebook news feeds across 22 months leading up to the election. The best-fitting linear and quadratic lines are drawn on the plots. For Whitman the percentage of first-person pronouns decreased significantly with the linear time predictor, b = −0.053, t(20) = −2.10, p < 0.049, and this predictor explained a significant proportion of the variance, R2 = 0.18, F (1,20) = 4.42, p < 0.049. Addition of the quadratic predictor did not increase the variance explained, b = 0, t(19) < 1.0, ns. For Brown the percentage of first-person pronouns also decreased significantly with the linear time predictor, b = −0.064, t(20) = −3.00, p < 0.007, and this predictor explained a significant proportion of the variance, R2 = 0.31, F (1,20) = 8.98, p < 0.007. Addition of the quadratic predictor marginally improved the prediction, b = 0.007, t(19) = 2.02, p < 0.06, and significantly increased the variance explained, R2 = 0.43, F (2,19) = 7.22, p < 0.005. For Whitman the percentage of third-person pronouns remained the same throughout the study period. For Brown, on the other hand, the percentage of third-person pronouns increased significantly with the linear time predictor, b = 0.039, t(20) = 2.04, p < 0.05, and this predictor explained a significant proportion of the variance, R2 = 0.17, F (1,20) = 4.17, p < 0.054. Addition of the quadratic predictor did not reach significance, b = −0.006, t(19) = −1.79, p < 0.09, but with the two predictors the proportion of variance explained was considerably increased, R2 = 0.29, F (2,19) = 3.91, p < 0.038.

223

4.2. Cognition and affect

224

Table 4 shows the mean number of words in the cognition (the “cognitive mechanism” category in LIWC) and affect categories for both candidates combined across the 22 time periods in this study. A 2 × 2 repeated-measures ANOVA was performed using time period as if it were a subject factor. There was significantly more use of terms related to cognition (mean = 15.08 percent) than affect (mean = 5.66 percent), F (1,21) = 2126.09, p < 0.0001, and significantly more terms of both types used by Whitman in contrast to Brown, F (1,21) = 19.30, p < 0.0001. However, there was also an interaction between the cognition/affect variable and the candidate F (1,21) = 24.44, p < 0.0001. Examination of the means suggests that although the percentage of affect terms was the same for both candidates (means = 5.66 percent for Brown and also for Whitman), the percentage of cognition terms was slightly higher for Whitman (mean = 15.94 percent) than for Brown (mean = 14.21 percent). Figure 3 shows the percentage of comments that were categorized by LIWC as being related to cognition and the number of comments that were categorized as being related to affect over the 22 months leading to the election. The best-fitting linear and quadratic lines are drawn on the plots. For Whitman the percentage of words related to cognitive mechanism decreased marginally with the linear time predictor, b = −0.051, t(20) = −1.99, p < 0.06, and this predictor explained a marginally

204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221

225 226 227 228 229 230 231 232 233 234 235 236 237 238

Galley Proof

8

10/05/2013; 9:21



Fig. 2. Linear and quadratic trends for the percentage of first-person and third-person verbs in the news feeds of Meg Whitman and Jerry Brown in the months leading up to the 2010 California Gubernatorial election. 239 240 241 242 243 244 245 246 247

significant proportion of the variance, R2 = 0.17, F (1,20) = 3.97, p < 0.06. Addition of the quadratic term produced a much better prediction, b = −0.011, t(19) = −2.99, p < 0.008, and considerably increased the variance explained, R2 = 0.43, F (2,19) = 7.24, p < 0.005. For Brown the percentage of words related to cognitive mechanism remained the same throughout the study period. For Whitman the percentage of words related to affect was not predicted by the linear time predictor, b = −0.024, t(20) = −0.84, ns, but addition of the quadratic term produced a significant prediction, b = 0.010, t(19) = 2.23, p < 0.038. Despite this, the proportion of variance explained by the two terms was not significant, R2 = 0.23, F (2,19) = 2.90, p < 0.079. For Brown the percentage of words related to affect remained the same throughout the study period.

Galley Proof

10/05/2013; 9:21



9

Table 5 Mean percentage of positive and negative sentiment words in two political candidates’ Facebook news feeds combined across 22 months preceding an election Sentiment category Positive sentiment Negative sentiment Mean

2010 California gubernatorial candidates Meg Whitman Jerry Brown 4.29 3.92 1.45 1.67 2.87 2.79

Mean 4.10 1.56

Fig. 3. Linear and quadratic trends for the percentage of “cognitive mechanism” and “affect” terms in the news feeds of Meg Whitman and Jerry Brown in the months leading up to the 2010 California Gubernatorial election.

Galley Proof

10

10/05/2013; 9:21



Fig. 4. Linear and quadratic trends for the percentage of positive and negative sentiment terms in the news feeds of Meg Whitman and Jerry Brown in the months leading up to the 2010 California Gubernatorial election. 248

4.3. Converging sentiment

249

Table 5 shows the mean number of words in the positive and negative affect categories of LIWC for both candidates combined across the 22 time periods in this study. A 2 × 2 repeated-measures ANOVA was performed using time period as if it were a subject factor. There was significantly more use of words reflecting positive affect (mean = 4.10 percent) than words reflecting negative affect (mean = 1.56 percent), F (1,21) = 296.88, p < 0.0001. There were no other significant effects. Figure 4 shows scatterplots of the percentage of positive and negative sentiment terms in the comments

250 251 252 253 254

Galley Proof

10/05/2013; 9:21



11

272

from both candidates’ Facebook news feeds across 22 months leading up to the election. The best-fitting linear and quadratic lines are drawn on the plots. For Whitman the percentage of words reflecting positive sentiment decreased significantly with the linear time predictor, b = 0.067, t(20) = −2.43, p < 0.025, and this predictor explained a significant proportion of the variance, R2 = 0.23, F (1,20) = 5.90, p < 0.05. Addition of the quadratic predictor did not increase the variance explained, b = −0.005, t(19) = 1.11, ns. For Brown the percentage of words reflecting positive sentiment was not significantly related to time using the linear predictor alone, however addition of the quadratic component resulted in a significant prediction, b = −0.012, t(19) = −2.48, p < 0.011 and explained a significant proportion of variance, R2 = 0.32, F (2,19) = 4.47, p < 0.05. For Whitman the percentage of words reflecting negative sentiment increased significantly with the linear time predictor, b = 0.043, t(20) = 3.44, p < 0.003, and this predictor explained a significant proportion of the variance, R2 = 0.37, F (1,20) = 11.86, p < 0.003. Addition of the quadratic predictor did not increase the variance explained, b = 0.004, t(19) = 1.77, ns. For Brown the percentage of words reflecting negative sentiment was not significantly related to time using the linear predictor alone, however addition of the quadratic component resulted in a significant prediction, b = −0.006, t(19) = −2.40, p < 0.027 and explained a (weakly) significant proportion of variance, R2 = 0.326, F (2,19) = 3.26, p < 0.061.

273

4.4. Addressee and content

274

Based on prior studies [60,62] and informal sorting data from subjects, the following six comment categories were coded: – Direct address to the candidate. ∗ “I’ll support you and I hope you support my 2nd amendment rights and be tough on immigration.” ∗ “America Needs More SOUL. I believe in you. Go Jerry Go!!!” – Direct address to another commenter ∗ “I think this lady has what it takes. Look at her history, she has been a success everywhere she’s been.” ∗ “This woman has nothing to say. Its all meaningless slogans.” – Evidence ∗ “Meg Whitman Just turned coat on immigration. She just put an ad on Spanish TV networks saying she was against Arizona’s 1070, and Proposition 187, A law that would have denied Illegal immigrants access to public benefits. TWO FACED!” ∗ “The New Jim Crow www.huffingtonpost.com There are more African Americans under correctional control today – in prison or jail, on probation or parole – than were enslaved in 1850, a decade before the Civil War began.” – Action ∗ “Should Illegal Aliens be Granted In-State (Resident) Tuition Rates at State Colleges and Universities? Please vote and leave your comments: http://www.stopillegalaliens.info/” ∗ “If you dont like Meg join this! http://www.facebook.com/home.php?#!/pages/I-bet-this-pencilcan-get-more-fans-than-Meg-Whitman/110010405703205” – Rebuttal

255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271

275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295

Galley Proof

12

10/05/2013; 9:21



Fig. 5. Number of comments in each comment category across four time periods from Jerry Brown’s Facebook wall.

Fig. 6. Number of comments in each comment category across four time periods from Meg Whitmans’s Facebook wall. 296 297 298 299 300 301 302 303 304 305 306 307 308 309

∗ “If the worse thing she’s done is not vote, shown the same apathy towards government that many of us has felt, while building a hugely successful business, I can live with that!” ∗ “This is getting old, there are more important things that should be on the court dockets then this! it’s time to let people live their lives the way they want as long as they don’t break any laws it’s not up to us to make this choice for them.” – Joking and Ridicule ∗ “I have a fabulous idea for the Nutmeg political action team. Just find a way to hook this page to the Hyperion Plant in LA (wastewater treatment plant). Presto! A ready stream of already steaming statements!” ∗ “I just found out that my 4th grade daughter has been representing the ‘Jerry Brown voters’ in her class discussions about the governors race. I just hope she hasn’t been repeating everything she’s heard me say about Meg Whitman, or I’ll be getting a note from her teacher soon.” The first two categories are particularly important with regard to the hypothesis that converging sentiment and reflection-to-selection involve a turn from inward-directed, community-building activity to

Galley Proof

10/05/2013; 9:21



13

Fig. 7. Number of comments directed toward the candidate, the opponent, and others across four time periods from Jerry Brown’s Facebook wall.

Fig. 8. Number of comments directed toward the candidate, the opponent, and others across four time periods from Meg Whitman’s Facebook wall. 310 311 312 313 314 315 316 317 318 319 320 321 322

more outward-directed activity. Comments in the first category are relevant to identity-based participation [57] since they involve interaction with the object around which communality is being constructed. Comments in the second category can reflect community-building if they are positive and directed towards others within the community or negative and directed to outsiders. They are disruptive to the community if they are negative and directed to community members or positive and directed to outsiders. Figures 5 and 6 show the number of comments in each category across the four time periods (out of 100 per time period) for the two candidates. The number of comments differed considerably across the categories, x2 (5) = 831.58, p < 0.01 and x2 (5) = 719.37, p < 0.01 for Brown and Whitman respectively, with direct address to candidates and direct address to other commenters dominating. In Brown’s case, direct address to the candidate decreased across time periods, x2 (5) = 8.85, p < 0.05. Figures 7 and 8 show comments combined across all of the previous categories and resorted in terms of whether they were directed toward the candidate, the opponent, or others. Again, for Brown com-

Galley Proof

14

10/05/2013; 9:21



Fig. 9. Number of comments in the five intention categories on the candidates’ Facebook walls.

344

ments directed at the candidate decreased over time, x2 (3) = 7.86, p < 0.05, while comments directed at others increased over time, x2 (3) = 13.40, p < 0.05. Although there was a similar trend for Whitman, differences across time periods were not significant. There were very few comments within either candidate’s site that were directed at the opponent. Direct address to the candidate constituted a large percentage of all comments. Users had many goals when they made a direct address. Categorization of comments within the direct address category was carried out focusing on the intention of the commenters. The following seven types of comments were identified: – Question: Asking the candidate for his or her position on an issue. – Suggestion: Telling the candidate that they should do or try something. – Request: Asking the candidate to do something. – Support: Expressions of enthusiasm and encouragement. – Invitation: Asking the candidate to participate in something. – Stories: Telling personal narratives or describing things that have happened to the poster. Figure 9 shows the number of comments in each intention subcategory for both candidates. Trends over time varied dramatically but non-systematically across time periods, so the data was collapsed over time. The distribution of comment types was unequal across categories and between candidates, x2 (7) = 36.7, p < 0.05. The most significant observations of differences between candidates were that Whitman was asked three times as many questions as Brown, but Brown received 7 times as many requests as Whitman. Brown also received considerably more support activity. Across categories, it is clear that one of the primary purposes of dialog on candidate’s websites is to offer support.

345

5. Discussion

346

In several instances over two elections we observed the reflection-to-selection phenomenon, which we initially described as a linear model in which commenters are focused inward when they begin to build their online community but turn outward with time. Our new analyses show that another pattern exists in which commenters return to an inward orientation at election time. This pattern was present

323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343

347 348 349

Galley Proof

10/05/2013; 9:21


S.P. Robertson et al. / Political discourse on social networking sites 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395

15

in the data for the winner of the California Gubernatorial race (Brown’s data in Fig. 2), but not in the loser’s data. It remains to be seen if this is replicable and, if so, explainable. It is interesting that the inward-facing first-person pronoun and the outward-facing third-person pronoun seem to trade off. The U-shaped function for first person pronoun data for Brown is mirrored by an inverted U-shape in the third-person pronoun data. These two pronoun types are theoretically independent of each other, they could go up or down together or be completely unrelated. However this effect suggests that commenters are focused predominantly one way or the other at any given time and may drift from one focus to the other as their goals change through time. In several instances over two elections we observed the converging sentiment phenomenon, which we also initially described as a linear model in which commenters are most positive and least negative when they are first forming their community, but become less positive and more negative over time. Our new analysis again shows that another pattern exists in which commenters may become more positive and less negative once again at the end of their quest as the event that they have organized around draws near. At least this pattern was present for the winner of the election. The community formed around the loser of the election showed a different pattern in which they plateaued at a lower level of positive commenting as the election neared, but accelerated the increase in comments reflecting negative affect. In data first presented here, we found that political discourse in SNSs consists of both rational discussion (as reflected by the cognitive mechanism category of LIWC) and affective commenting (as reflected in the affect category of LIWC). Rational discussion is more prevalent than affective commenting, and this trend remains the same over at least several months prior to the election. Whitman’s data showed a tantalizing trend in which at first rational commenting increased through time and affecting commenting decreased through time, but as the election neared this trend reversed such that rational commenting began to decline and affective commenting began to increase. Like pronoun direction and affective valence, these two types of discourse are independent and do not have to track each other reciprocally. The observation that they do suggests that communities cannot simultaneously maintain both rational and affective interaction. Perhaps a type of affect contagion, or rationality contagion, takes place to move the whole community into one discourse context or the other. The direction of movement of a community along these scales may say a lot about the health of the community at any given moment in time. Content analysis of the comments showed a clear trend such that engagement with a community’s candidate is dominant early in campaigns and engagement with others is dominant in later stages. One explanation for this trend is that political communities built around candidates need to construct and solidify the components of their common interest. This means initially asking many questions of the candidate. This period tends to be highly positive in terms of comment sentiment. After the group is more well defined, however, the communal interest becomes focused on the other candidate and the competitive group(s). Community members evolve from a reflective mode into a more active mode that is sensitive to external events and the ongoing contest. There becomes more of an emphasis on defining the other, which in the context of politics generates more negative comments. The predominance of cognitive words in the discourse on candidates’ Facebook pages is good news for proponents of the view that SNSs can serve as a public sphere. Since one of the defining characteristics of a public sphere is that it allows rational arguments to be aired and discussed openly, the presence of these terms suggests that this activity is indeed taking place with high frequency. Of course, the presence of affective terms also shows that public discourse on SNSs is passionate. While these types of discourse may fluctuate in relative frequency, they are co-present at all times. Readers browsing SNS material must skip through many types of comments and posts in order to experience the content. In this way SNSs are unlike face-to-face public spheres in which attention must be managed more carefully and participants are not free to navigate the information threads of their own interests.

Galley Proof

16 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440

10/05/2013; 9:21



SNS’s might better be described as user-generated “potential public spheres”, information-opportunity contexts that provide options for browsing, navigation and participation. They become “realized public spheres” only when people take action, for example by successfully finding interesting information, making reasoned decisions, contributing comments, and/or translating their experience into the real world as a vote or other political action. This conceptualization of the public sphere changes it from a sociological phenomenon that can be observed and measured from the outside to a psychological or experiential phenomenon that depends on participants’ states of mind and actions. It means that even when several people are in the same virtual space (like a SNS) some may be considered to be in a public sphere while others are not. The affordances of SNSs can facilitate the public sphere envisioned by Habermas [7,8,20] as citizens can gather together virtually, irrespective of geographic location, and engage in information exchange and rational discussion as long as they have an Internet connection. However, only when the technology is used in context does its utility emerge as a resource for political deliberation. Our study has shown that the “potential public spheres” available through social media were enacted, in turn, becoming “realized public spheres.” Many people may be engaging politically in various online public spheres without contributing content through lurking. Others may be more visible in the public spheres available online through SNSs. Whereas in the physical world we typically have awareness of those who are taking part in a discussion, SNSs enable users to peruse persistent discussions over time (as long as they are not deleted). The people who are posting information, opinions and contributing to discussions, may be making the more visible contributions, but there is an invisibility to effects of the contributions being made online, as many people can be effected by what is being posted without contributing themselves. This conceptualization of the public sphere opens up several possibilities for research and it will be a challenge for researchers to understand how or measure when public spheres move from potential to realized. We aim to further investigate the degree to which SNSs become realized public spheres by employing ethnographic and survey methods to better understand both what is visible online, and what is invisible (going on through people’s minds). The social network site of a politician during a campaign is a community space that is bounded in time and centered on a set of ideas, a common goal, and a defining event. Participants come to these spaces with multiple goals of their own, and this research shows that the distributions of these goals vary considerably for supporters of the different candidates and across time. The discrepancy in distribution of some types of comment categories between communities with the same purpose is potentially meaningful to people involved in those communities. For example, community members among Whitman’s supporters might have benefitted from knowing that she was being asked an inordinate number of questions relative to her competitor and that questions were not tailing off as might be expected from a community that was successfully establishing its shared interests. We observed, for example, that many of the questions Whitman was asked initially were clarification questions about her positions. She was not as well defined within her voting community as Brown, who was well known before the gubernatorial race. However, later questions began to be more critical, centering on Whitman’s perceived unresponsiveness within her Facebook community. We conjecture that the weaker trend of turning away from engagement with the candidate over time for Whitman, and her large number of questions relative to Jerry Brown, are related. To the degree that Meg Whitman’s community was not able to find common ground and establish a strong interest bond, they continued to be focused inward and maintain an information-gathering mode. Similarly, the Brown camp might have been interested to know that their candidate was receiving many more requests than his competitor. Requests imply an assumption of agency, that the person being asked is, or will be, in a position to fulfill the request. Similarly, a comparative analysis of supportive statements

Galley Proof

10/05/2013; 9:21


S.P. Robertson et al. / Political discourse on social networking sites 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483

17

would have revealed that Brown was receiving considerably more support than Whitman, even though there was more activity on Whitman’s site. Again we conjecture that the more clearly defined community interests allowed the group to turn their attention outwards as the election approached in the same manner that was observed in the 2008 presidential election [60,61]. Ren et al. [57] point out that choices made by designers of community and social systems influence the types of groups that can form and flourish within these systems. The Facebook comment area in 2010 was a general enough platform to support multiple goals and activities. However, Facebook and other SNSs are currently engaged in rapid redesign processes to support a myriad of different activities and user goals, often concurrently. This research suggests that focusing these efforts on artifacts, such as creating new picture-viewer or music-sharing features, is only one avenue of opportunity for designers. Another is to focus redesign efforts on support for different user goals, for example asking questions, making requests, telling stories, and expressing support. Some of these goals are general across different community contexts, whereas others might be context specific (e.g. expressing support is more likely to occur in the political arena, but asking questions will be a common activity for many groups). The potential information to be gleaned from comparative analysis of categories between interestbased groups with the same overarching interest (electing a candidate and understanding the issues), but with opposed subordinate goals (i.e. in competition) also offers a design space for human-computer interaction. In politics, campaigns could benefit from contrasting sentiment, question, request, and other categories across the SNSs of their competitors. In contexts outside of politics, contrasting patterns could be used to gauge the health or evolutionary stage of a community’s development. Finally, community members themselves could benefit from the design of social browsers that are sensitive to comment categories and user intentions. A new community member might, for example, be interested in browsing all of the questions that have been asked within the SNS in order to understand the issues better. Alternatively, a user might be interested in seeing all critical statements, or finding out how responsive a candidate or other members had been to requests. A “social affect browser” might be able to group comments in terms of sentiment and show users all of the angry comments or all of the supportive comments. A “social intention browser” might be able to group comments into motive categories such as questions, requests, and stories. Users would then be able to filter or navigate through different intentions. Admittedly, such judgments pose a difficult problem of linguistic analysis, but current interfaces to SNSs like Facebook, which focus mainly on chronology or friend-network structure to support browsing and navigation, are difficult to negotiate and fail to support many potential community-building activities. On the other hand, social browsers sensitive to sentiment and intention might also pose ethical challenges for end users and designers. For example, outside of the public realm of politics, it would be possible to determine if the comments on an individual’s SNS were highly negative relative to his/her friends. SNSs are the new public sphere for civic engagement. It is critical to understand how civic SNSs should evolve over time and to understand patterns that might indicate potential problems requiring attention. One of the core, shared interests of a civic community is the overall functional health of the community itself and its members. Information exists in the comment corpus of civic SNSs that could provide clues to the health and concerns of the community and that provide design opportunities for building and maintaining community through sentiment- and intention-based navigation, searching, and browsing.

Galley Proof

18

10/05/2013; 9:21



484

Acknowledgments

485

487

This research was funded by NSF award No. IIS-1064852 to the first author. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

488

References

486

489 490

[1] [2]

491 492

[3]

493 494 495

[4]

496 497 498

[5] [6]

499 500 501

[7] [8]

502 503

[9]

504 505

[10]

506 507

[11]

508 509

[12]

510 511

[13]

512 513 514 515 516

[14] [15] [16] [17]

517 518 519

[18]

520 521 522

[19] [20]

523 524

[21]

525 526 527

[22]

528 529

[23]

530 531

[24]

532 533

[25]

534 535 536

[26]

P. Agre, Real-time politics: The Internet and the political process, The Information Society 18(5) (2002), 311–331. S. Albrecht, Whose voice is heard in online deliberation? A study of participation and representation in political debates on the Internet, Information, Communication & Society 9(1) (2006), 62–82. J. Arguello, B.S. Butler, E. Joyce, R. Kraut, K.S. Ling, C. Rosé and X. Wang, Talk to me: Foundations for successful individual-group interactions in online communities, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’06), ACM, New York, NY, 959–968, 2006. M.D. Barton, The future of rational-critical debate in online public spheres, Computers and Composition 22 (2005), 177–190. B. Bimber and R. Davis, Campaigning online: The Internet in U.S. elections, Oxford: Oxford University Press, 2003. P. Boeder, Habermas’ heritage: The future of the public sphere in the network society, First Monday [Online] 10(5), (5 September 2005). C. Calhoun, ed., Habermas and the public sphere, Cambridge, Mass: MIT Press, 1992. C. Calhoun, Information technology and the international public sphere, in: Shaping the Network Society: The New Role of Civil Society in Cyberspace, D. Schuler and P. Day, eds, Cambridge: The MIT Press, 2004, pp. 229–251. D.B. Carlin, D. Schill, D.G. Levasseur and A.S. King, The post-9/11 public sphere: Citizen talk about the 2004 presidential debates, Rhetoric and Public Affairs 8(4) (2005), 617–638. L. Dahlberg, Computer-mediated communication and the public sphere: A critical analysis, Journal of ComputerMediated Communication [Online] 7(1) (2001), Available at http://jcmc.indiana.edu/vol7/issue1/dahlberg.html. L. Dahlberg, The Internet and democratic discourse: Exploring the prospects of online deliberative forums extending the public sphere, Information, Communication, & Society 4(4) (2001), 615–633. P. Dahlgren, The Internet, public spheres, and political communication: Dispersion and deliberation, Political Communication 22(2) (2005), 147–162. U.M. Dholakia, R.P. Bagozzi and L.K. Pearo, A social influence model of consumer participation in network- and smallgroup-based virtual communities, International Journal of Research in Marketing 21(3) (2004), 241–263. J. Donath and D. boyd, Public displays of connection, BT Technology Journal 22(4) (2004), 71–82. H. Farrell and D.W. Drezner, The power and politics of blogs, Public Choice 134 (2008), 15–30. K.A. Foot and S.M. Schneider, Web campaigning, Cambridge, MA: MIT Press, 2006. K.A. Foot, S.M. Schneider, M. Dougherty, M. Xenos and E. Larsen, Analyzing linking practices: Candidate sites in the 2002 US Electoral Web Sphere, Journal of Computer Mediated Communication [Online] 8(4) (2003), Available at http://jcmc.indiana.edu/vol8/issue4/foot.html. E. Gilbert, T. Bergstrom and K. Karahalios, Blogs are echo chambers: Blogs are echo chambers, Proceedings of the 42nd Hawaii International Conference on System Sciences, (HICSS ’09) (2009), 1–10, doi: 10.1109/HICSS.2009.91. J. Glaser and P. Salovey, Affect in electoral politics, Personality and Social Psychology Review 2(3) (1998), 156–172. J. Habermas, The structural transformation of the public sphere: An inquiry into a category of bourgeois society, Cambridge, Mass, MIT Press, 1989. K.N. Hampton, L.S. Goulet, L. Raine and K. Purcell, Social networking sites and our lives, Pew Internet and American Life Project (16 June 2011), Available at: http://pewinternet.org/Reports/2011/Technology-and-social-networks.aspx, Accessed on Sept. 20, 2011. J.T. Hancock, C. Landrigan and C. Silver, Expressing emotion in text-based communication, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’07), ACM, New York, NY, (2007), 929–932. R. Hastie, A primer of information-processing theory for the political scientist, in: Political Cognition: The 19th Annual Carnegie Symposium on Cognition, R.L. Richard and D.O. Sears, eds, Erlbaum, Hillsdale, NJ, (1986), 11–39. I. Himelboim, Civil society and online political discourse: The network structure of unrestricted discussions, Communications Research 38(5) (2011), 634–659. J. Horrigan, K. Garrett and P. Resnick, The internet and democratic debate, http://www.pewinternet.org/pdfs/PIP_ Political_Info_Report.pdforg/pdfs/PIP_Political_Info_Report.pdf, 2004. A.N. Joinson, ‘Looking at’, ‘looking up’ or ‘keeping up with’ people? Motives and uses of Facebook, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’08), ACM, New York, NY, (2008), 1027–1036.

Galley Proof

10/05/2013; 9:21


S.P. Robertson et al. / Political discourse on social networking sites 537 538

[27] [28]

539 540

[29]

541 542 543

[30]

544 545

[31]

546 547 548

[32]

549 550 551

[33]

552 553

[34]

554 555

[35]

556 557

[36]

558 559

[37]

560 561 562 563

[38] [39]

564 565

[40]

566 567

[41]

568 569

[42]

570 571

[43]

572 573

[44]

574 575

[45]

576 577

[46]

578 579

[47]

580 581 582

[48] [49]

583 584

[50]

585 586

[51]

587 588

[52]

589 590

[53]

591 592 593 594

[54]

19

B.D. Jones, Bounded rationality, Annual Review of Political Science 2 (1999), 297–321. A.D. Kramer, An unobtrusive behavioral model of ‘Gross National Happiness’, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’10), ACM, New York, NY, (2010), 287–290. A.D. Kramer, L.M. Oh and S.R. Fussell, Using linguistic features to measure presence in computer-mediated communication, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’06), ACM, New York, NY, (2006), 913–916. A.D. Kramer and K. Rodden, Applying a user-centered metric to identify active blogs, in: Extended Abstracts of the SIGCHI Conference on Human Factors in Computing Systems (CHI’07), ACM, New York, NY, (2007), 2525–2530. A.D. Kramer and K. Rodden, Word usage and posting behaviors: Modeling blogs with unobtrusive data collection methods, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’08), ACM, New York, NY, (2008), 1125–1128. C. Lampe, N. Ellison and C. Steinfield, A face (book) in the crowd: Social searching vs. social browsing, in: Proceedings of the 2006 20th Anniversary Conference on Computer Supported Cooperative Work, ACM, New York, NY, (2006), 167– 170. C. Lampe, R. Wash, A. Valesquez and E. Ozkaya, Motivations to participate in online communities, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’10), ACM, New York, NY, (2010), 1927–1936. R. Lau, Models of decision making, in: Oxford Handbook of Political Psychology, D. Sears, L. Huddy and R. Jervis, eds, Oxford University Press, Oxford, 2003. M. Lodge, K. McGraw and P. Stroh, An impression-driven model of candidate evaluation, American Political Science Review 83 (June 1989), 399–419. M. Lodge, M. Steenbergen and S. Brau, The responsive voter: Campaign information and the dynamics of candidate evaluation, American Political Science Review 89 (June 1995), 309–326. M. Madden and K. Zickuhr, 65% of online adults use social networking sites, Pew Internet and American Life Project (26 July 2011), Available at: http://pewinternet.org/Reports/2011/Social-Networking-Sites.aspx, Accessed on Sept. 20, 2011. G.E. Marcus, Emotions in politics, Annual Review of Political Science 3 (2000), 221–250. G.E. Marcus, W.R. Neuman and M. MacKuen, Affective intelligence and political judgment, University of Chicago Press, Chicago, 2000. C. Mascaro and S. Goggins, Brewing up citizen engagement: The coffee party on Facebook, Communities and Technologies 2011, Brisbane, Australia, (29 June–2 July 2011). C. Mascaro and S. Goggins, Twitter as virtual town square: Citizen engagement during a nationally televised republican primary debate, APSA 2012 Annual Meeting, New Orleans, LA, (30 August–2 September 2012). C. Mascaro, A. Novak and S. Goggins, The daily brew: The structural evolution of the coffee party on facebook during the 2010 united states midterm election season, Journal of Information Technology and Politics 9(3) (2012), 234–253. M. McPherson, L. Smith-Lovin and J. Cook, Birds of a feather: Homophily in social networks, Annual Review of Sociology 27(1) (2001), 415–444. M. Milliken, K. Gibson and S. O’Donnell, User-generated video and the online public sphere. Will YouTube facilitate digital freedom of expression in Atlantic Canada? American Communication Journal 10(3) (2008), 1–14. M. Milliken and S. O’Donnell, User-generated online video: The next public sphere? IEEE International Symposium on Technology and Society (2008), 1–3. N.H. Nie, Sociability, interpersonal relations, and the Internet: Reconciling conflicting findings, American Behavioral Scientist 45(3) (2001), 426–437. Nielsen, State of the media: The social media report Q3, 2001, Available at: http://www.nielsen.com/content/dam/ corporate/us/en/reports-downloads/2011-Reports/nielsen-social-media-report.pdf, Accessed on Sept. 20, 2011. Z. Papacharissi, The virtual sphere, New Media and Society 4(1) (2002), 9–27. Z. Papacharissi, Democracy online: Civility, politeness, and the democratic potential of online political discussion groups, New Media & Society 6(2) (2004), 259–283. J.W. Pennebaker, M.E. Francis and R.J. Booth, Linguistic Inquiry and Word Count: LIWC (2nd Edition), Mahwah, NJ: Lawrence Erlbaum Associates, 2001. J.W. Pennebaker, M.R. Mehl and K.G. Niederhoffer, psychological aspects of natural language use: Our words, our selves, Annual Review of Psychology 54 (2003), 547–577. N. Poor, Mechanisms of an online public sphere: The website Slashdot, Journal of Computer-Mediated Communication [Online] 10(2) (2005), article 4. Available at http://jcmc.indiana.edu/vol10/issue2/poor.html. T.R. Postmes and R. Spears, Refining the cognitive redefinition of the group: Deindividuation effects in common-bond vs. common-identity groups, in: Side Effects Centre Stage: Recent Developments in Studies of Deindividuation in Groups, T. Postmes, R. Spears, M. Lea and S. Reicher, eds, Amsterdam: KNAW, 2000, pp. 63–78. D.A. Prentice, D.T. Miller and J.R. Lightdale, Asymmetries in attachments to groups and to their members: Distinguishing between common-identity and common-bond groups, Personality and Social Psychology Bulletin 20(5)

Galley Proof

20 595 596

[55]

597 598 599

[56]

600 601

[57]

602 603 604

[58] [59]

605 606

[60]

607 608 609

[61]

610 611

[62]

612 613

[63]

614 615 616 617

[64] [65]

618 619 620

[66] [67]

621 622 623

[68] [69]

624 625 626 627

[70] [71] [72]

628 629

[73]

630 631

[74]

632 633 634

[75] [76]

635 636 637 638

[77] [78]

639 640

[79]

641 642

[80]

10/05/2013; 9:21


S.P. Robertson et al. / Political discourse on social networking sites (1994), 484–493. A. Quan-Haase, B. Wellman, J. Witte and K. Hampton, Capitalizing on the Internet: Social contact, civic engagement, and sense of community, in: Internet and Everyday Life, B. Wellman and C. Haythornthwaite, eds, Oxford: Blackwell, 2002, pp. 291–324. D.P. Redlawsk, You must remember this: A test of the on-line model of voting, The Journal of Politics 63(1) (2001), 29–58. Y. Ren, R. Kraut and S. Kiesler, Applying common identity and bond theory to design of online communities, Organization Studies 28(3) (2007), 377–408. W. Riker and P. Ordeshook, A theory of the calculus of voting, American Political Science Review 62(1) (1968), 25–42. S.P. Robertson, Changes in referents and emotions over time in election-related social networking dialog, in: Proceedings of the 44th Annual Meeting of the Hawaii International Conference on Systems Sciences (HICSS44), IEEE, 2001, 10pp. S.P. Robertson, R.K. Vatrapu and R. Medina, The social life of social networks: Facebook linkage patterns in the 2008 U.S. presidential election, in: Proceedings of the 10th International Digital Government Research Conference (dg.o ’09), ACM, New York, NY, USA, (2009), 6–15. S.P. Robertson, R.K. Vatrapu and R. Medina, Off-the-wall political discourse: Facebook use in the 2008 U.S. presidential election, Information Polity 15(1–2) (2010), 11–31. S.P. Robertson, R.K. Vatrapu and R. Medina, Online video “friends” social networking: Overlapping online public spheres in the 2008 U.S. presidential election, Journal of Information Technology and Politics 7(2–3) (2010), 182–201. S.P. Robertson, S. Douglas, M. Maruyama and L. Chen, Political dialog evolution in a social network, in: Proceedings of the 13th Annual International Conference on Digital Government Research (dg.o ’12), ACM, New York, NY, USA, (2012), 40–48. H.A. Simon, Information processing models of cognition, Annual Review of Psychology 30 (1979), 363–396. H.A. Simon, Human nature in politics: The dialogue of psychology with political science, American Political Science Review 79 (1985), 293–304. H.A. Simon, Rationality in political behavior, Political Psychology 16 (1995), 45–61. J. Stromer-Galley, Diversity of political conversation on the Internet: Users’ perspectives, Journal of Computer-Mediated Communication [Online] 8(3) (2003), Available at http://jcmc.indiana.edu/vol8/issue3/stromergalley.html. N.J. Stroud, Media effects, selective exposure, and Fahrenheit 9/11, Political Communication 24(4) (2007), 415–432. N.J. Stroud, Media use and political predispositions: Revisiting the concept of selective exposure, Political Behavior 30 (2008), 341–366. M. Tremayne, Blogging, citizenship, and the future of media, New York: Routledge, Taylor & Francis Group, 2006. S. Utz, Social identification and interpersonal attraction in MUDs, Swiss Journal of Psychology 62(2) (2003), 91–101. S. Utz and K. Sassenber, Disruptive justice in common-bond and common-identity groups, Group Processes and Intergroup Relations 5(2) (2002), 151–162. R.K. Vatrapu, S.P. Robertson and W. Dissanayake, Are political weblogs public spheres or partisan spheres? International Reports on Socio-Informatics 5(1) (2008), 7–26. K. Wallsten, Many sources, one message: Political blog links to online videos during the 2008 campaign, Journal of Political Marketing 10(1–2) (2011), 88–114. B. Warnick, Rhetoric online: Persuasion and politics on the world wide web, New York: Peter Lang Publishing, 2007. B. Wellman, A. Quan-Haase, J. Witte and K.N. Hampton, Does the Internet increase, decrease, or supplement social capital? Social networks, participation, and community commitment, American Behavioral Scientist 45(3) (2001), 437– 456. A. Wilhelm, Democracy in the digital age: Challenges to political life in cyberspace, New York: Routledge, 2000. C. Williams and G. Gulati, Social networks in political campaigns: Facebook and the 2006 midterm elections, Annual Meeting of the American Political Science Association, Chicago, Illinois, (2007). C. Williams and G. Gulati, The political impact of Facebook: Evidence from the 2006 midterm elections and 2008 nomination contest, Politics and Technology Review 1 (2008), 11–21. W. Wirth and H. Schramm, Media and emotions, Communication Research Trends 24(3) (2005), 3–39.