Making retweeting social: The influence of content and context information on sharing news in Twitter

Computers in Human Behavior 46 (2015) 75–84

Contents lists available at ScienceDirect

Computers in Human Behavior journal homepage: www.elsevier.com/locate/comphumbeh

Making retweeting social: The influence of content and context information on sharing news in Twitter Anja Rudat ⇑, Jürgen Buder 1 Knowledge Media Research Center, Schleichstr. 6, 72076 Tübingen, Germany

a r t i c l e

i n f o

Article history:

Keywords: Agent awareness Informational value News Retweeting Social navigation Twitter

a b s t r a c t Spreading news in Web 2.0 is easy and ubiquitous, especially in Twitter via retweeting. However, while some news develops viral power, other remains disregarded. The paper presents two laboratory experiments about potentially influencing criteria on retweeting. Study 1 investigated whether content criteria (informational value) and context criteria (agent awareness) influence retweeting decisions. It was hypothesized that agent awareness would moderate the influence of informational value on retweeting. Results did not confirm the hypothesis but instead revealed that both high informational value and agent awareness information led to retweeting. Further, the influence of both informational value and agent awareness on retweeting was mediated by the perceived importance of the tweets. Study 2 investigated the influence of the informational value of the news, agent awareness, and in-group reference. It was hypothesized that the influence of informational value and of agent awareness on retweeting are moderated by in-group reference. The results confirmed these assumptions and showed that informational value had more influence if agents did not belong to a salient in-group compared to if they did. In contrast, agent awareness had more influence if agents belonged to a salient in-group compared to if they did not. Ó 2015 Elsevier Ltd. All rights reserved.

1. Making retweeting social: The influence of content and context information on sharing news in Twitter News is an integral part of daily life: Among other things, news gives information and orientation, builds and shapes public opinion, helps people to reduce uncertainty and helps to create impressions of the world. News is an object of discussions among housewives as well as among politicians (Shoemaker, 2006). Therefore, it is an important trend that nowadays, Web 2.0 and social media applications are much more than just networks for connecting with old friends or meeting new people with similar interests. More and more, Web 2.0 services understand themselves as a kind of news media (e.g., Kwak, Lee, Park, & Moon, 2010) allowing their users sharing latest news with other people, and thus, taking part in a process that formerly was dedicated only to journalists (Hermida, 2010). This is part of the most prominent features that Web 2.0 brought along: the switch from only reception to participation of the users (O’Reilly, 2005). Lee and Ma (2012) stated that ‘‘sharing news in social media [has] become a phenom⇑ Corresponding author. Tel.: +49 7071 979 326; fax: +49 7071 979 100. E-mail addresses: [email protected] (A. Rudat), [email protected] (J. Buder). 1 Tel.: +49 7071 979 326; fax: +49 7071 979 100. http://dx.doi.org/10.1016/j.chb.2015.01.005 0747-5632/Ó 2015 Elsevier Ltd. All rights reserved.

enon of growing social, economic, and political importance’’ (p. 331). Twitter is one of the most frequently used Web 2.0 applications for sharing news (Kwak et al., 2010; Zarella, 2009). On Twitter, users can easily write, read, and share short messages, socalled tweets. Because of its viral power (Hansen, Arvidsson, Nielsen, Colleoni, & Etter, 2011) and its potential to touch the masses, Twitter has been object of research in many respects. For example, in one of the first studies on Twitter, Java, Song, Finin, and Tseng (2007) found that sharing information and news are named among the main intentions for using Twitter. This might have been a first indication of the potential that Twitter has as a news and information source. A feature of Twitter that makes spreading news quick and easy is retweeting. Retweeting means to forward a tweet to other users, namely, the followers. Followers are those people in Twitter who are subscribed to accounts of other users and receive their messages. Retweeting can be done by copying the respective tweet and adding ‘‘RT’’ to it or just by a simple mouse click. The latter way might be the one which made the idea of spreading information and news so popular. As retweeting is one of the mechanisms being responsible for the phenomena of virality and real-time information, it has been a topic of interest to researchers with respect to various aspects of retweeting motivations and determinants. Most studies investigated either content-related or

76

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

context-related factors, that is, either features of the message itself or characteristics of the person or the environment. Regarding the first, content-related factors, a content analysis by Kwak et al. (2010) reported that among the most retweeted tweets news made up a substantial part. They thus concluded that Twitter might be regarded as a news medium. Zarella (2009) came to a similar result and found that news was retweeted very often. With respect to sentiments as a factor influencing retweeting decisions, research yielded inconclusive results. For example, Hansen et al. (2011) found that negative news has more viral potential than positive news. In contrast, with regard to online news, Berger and Milkman (2012) showed that positive news was actually more viral. Regarding the second, context-related factors, Boyd, Golder, and Lotan (2010) studied retweeting from a conversational perspective. In case studies in which they asked Twitter users for their retweeting motivations, they found that reaching new audiences, entertainment, and seeking for validation were among the most frequently mentioned reasons for retweeting (Boyd et al., 2010). Further, the number of followers seems to positively affect retweeting decisions, as Suh, Hong, Pirolli, and Chi (2010) found in an analysis of a large sample of field data. However, the number of followers was also found to have a curvilinear effect on source credibility, such that having too many or too few followers leads to lower judgments of expertise and trustworthiness (Westerman, Spence, & Van Der Heide, 2012). Thus, the actual effect of source credibility on retweeting is still an issue of further research. In this paper, we aim to extend the research on possible content and context factors that might influence users’ decisions to retweet news. To shed more light on the question what makes some kind of news to be spread more likely than other, we present two experimental studies. Both studies address content and context factors that might impact sharing news in Twitter. Regarding content factors, we investigate the influence of informational value of the news, a concept we derived and adapted from news value theory (Galtung & Ruge, 1965). We focus on this content factor as news value theory provides a systematically developed typology of news content’s characteristics (Maier, Stengel, & Marschall, 2010). Moreover, news value theory is well established in the research field of news selection, which we are also interested in. Regarding context factors, we argue that in Web 2.0 applications, such as Twitter, other people play an important role. Users have to take care of their audience, thus they are likely to attend to social and contextual cues that are available in their environment. How social cues in computer-mediated environments can shape behavior, is at the heart of research on (group-) awareness (Janssen & Bodemer, 2013). Research questions in this field address how cognitive and/or behavioral traces of users can impact reactions of other users. As using Twitter is a computer-mediated communication setting in which users’ behavior can influence other users’ behavior we refer to research on awareness and investigate the influence of information about characteristics of other users on individual retweeting behavior (agent awareness). In the following chapters, both informational value and agent awareness will be explained and discussed. 2. The influence of the content on retweeting: Informational value News value theory is one of the approaches within the research tradition of news selection, focusing on the relevant characteristics of events, namely, news factors that are ascribed by journalists and that influence journalists’ decisions to report about the events in question (Galtung & Ruge, 1965; Lippmann, 1922). Research has further developed news value theory, and extended the scope of this theory by refining the list of news factors (Eilders, 2006;

Harcup & O’Neill, 2001; Papacharissi & de Fatima Oliveira, 2012; Rosengren, 1974; Rössler, Bomhoff, Haschke, Kersten, & Müller, 2011; Sande, 1971; Staab, 1990). Among other things, it was shown that the value of news influences not only journalists but also recipients in their selection decisions for consumption (Eilders, 1997; Eilders & Wirth, 1999). In the current paper, we focus on eight news factors (see Table 1) that have turned out to be meaningful and useful to be adapted to the notion of informational value. Informational value (Rudat, Buder, & Hesse, 2014) as an adapted concept subsumes these eight news factors into two groups. The first group contains those news factors that have in common that they either affect a large audience and/or have the potential to impact the audience’s mind or behavior. In contrast, the second group contains those news factors that do neither the one nor the other. In this regard, Relevance is a factor with high informational value as it should make a message more meaningful to a large audience. Moreover, Controversy, Negative Consequences, and Unexpectedness should also yield high informational value as these are types of information that can change one’s own and a recipient’s mind (positioning oneself in a controversy; thinking about how one could escape from negative consequences; integrating an unexpected event into one’s mental model). In contrast, our rationale classifies Proximity, Prominence, Personalization, and Aggression as news factors with low informational value. Proximity, by our definition, should have an immediate impact for only a part of one’s audience. Prominence, Personalization, and Aggression are factors that do not have immediate behavioral implications for recipients. In earlier experimental studies using systematically prepared fictive tweets about news topics (Rudat, 2014), the construct of informational value and its influence on retweeting decisions was tested. In an online experiment, results confirmed that informational value is associated with the two suggested concepts: (1) affecting a large audience and (2) having the potential to restructure the minds of recipients or to evoke behavioral change of recipients. Then, in a laboratory experiment, results showed that high informational value leads to more retweeting than low informational value. For a more detailed explanation of the concept see also (Rudat, 2014) and (Rudat et al., 2014). Since we found the influence of informational value on retweeting to be a stable effect, we aim to further investigate whether context-related factors might moderate it. Therefore, in the following, we will discuss awareness information about agents that should lead to social navigation.

3. The influence of the context on retweeting: Agent awareness In the context of social media and Web 2.0 not only content factors should influence selection decisions, but also characteristics and behavior of other people. Therefore, we draw on research on awareness (e.g., Janssen & Bodemer, 2013). Awareness in general means the state of consciousness about for example, other people, objects, feelings, or conditions (e.g., Carroll, Neale, Isenhour, Rosson, & McCrickard, 2003). Establishing awareness in computer-mediated settings can be facilitated by means of providing so-called awareness information, for example, about opinions, knowledge or preferences of interaction partners. A large number of studies could already show that awareness information does have a positive effect on efficient communication behavior or learning (Buder, 2011; Buder & Bodemer, 2008; Engelmann, Dehler, Bodemer, & Buder, 2009; Janssen & Bodemer, 2013; Janssen, Erkens, & Kirschner, 2011; Phielix, Prins, Kirschner, Erkens, & Jaspers, 2011; Schreiber & Engelmann, 2010). We differentiate between two kinds of awareness information: First,

77

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

Table 1 Alphabetically sorted list of news factors, their meaning (cf. Ruhrmann & Göbbel, 2007; Translation by the Authors) and psychological functions related to selection decisions of Twitter users. News factor

Meaning

Large audience and/ or impact on the audience

Aggression Controversy Negative consequences Personalization

The message is about threatened or practiced violence The message explicitly presents differences of opinions The possible or actual negative consequences of events are explicitly mentioned in the message

No Yes Yes

Individuals get a special meaning within an event in the message. One person or a few people are illustrated or even portrayed standing for a group or a company The message is about a popular person, popularity regardless of his or her actual political/economic power The message is about an event within a short geographical distance The message contains an event or a development that does already or will directly affect a large number of people The message is about an event or a development that cannot be predicted or stands in contrast to existing expectations

No

Prominence Proximity Relevance Unexpectedness

awareness information about the audience and second, awareness information about agents. Regarding the first kind of awareness, information about the audience can be made salient. This could lead to audience design (Clark & Murphy, 1982), that is, adapting one’s communication behavior to the audience’s properties. In a prior study (Rudat et al., 2014), we could show that information about characteristics of the user’s audience leads to audience design as it influenced the participants’ retweeting decisions according to the audience’s interests. Moreover, participants who were aware that a part of their audience (i.e., their followers) is interested in a specific topic exhibited more adaptation of their retweeting behavior than participants who were made aware of relatively unspecific information (gender distribution of the audience). However, the results of the study also revealed that high informational value was still preferred for retweeting over low informational value. Hence, audience awareness could not moderate the effect of informational value. Therefore, we introduce and test a second kind of awareness information, namely, agent awareness. Agents are other Twitter users who also retweet and thus participate in the process of spreading news. This means that information about the retweeting behavior of agents could be collected and aggregated, and fed back to an individual user. Being aware of what other agents already have done, should result in social navigation (Dourish & Chalmers, 1994; Höök, Benyon, & Munro, 2003). Social navigation means that people leave ‘‘footprints in the snow’’ (Höök et al., 2003, p.1), which are social cues that help other people to make their decisions about where to go and what to choose. The phenomenon of social navigation is probably well known to every Internet user: For instance, users receive recommendations about what to look at or what to purchase, based on their navigation patterns, their purchasing behaviors, or the evaluations of other users; online newspapers rank the most commented or most viewed articles; in online forums the best posts or answers are marked with a five star rating, indicating that many other people found a particular post relevant or helpful. Symbols such as little star icons may indicate quality, attractiveness, or popularity of an item and are therefore likely to influence users’ selection decisions (Winter, Krämer, Appel, & Schielke, 2010). On a psychological level, social navigation can be linked to the bandwagon effect (e.g., Sundar, Oeldorf-Hirsch, & Xu, 2008). A user should behave according to the behavior of many people before. This is based on the social heuristic that many other people probably were right (Axsom, Yates, & Chaiken, 1987; Sundar, Xu, & Oeldorf-Hirsch, 2009). Such a heuristic has the advantage of reduced cognitive effort that might be associated with selection decisions in the retweeting context (Fu, 2012; Shah & Oppenheimer, 2008). Although the bandwagon effect has been often discussed in the context of (political) opinion formation (e.g., Nadeau, Cloutier, & Guay, 1993), research on social navigation mechanisms has also employed the bandwagon

No No Yes Yes

heuristic as a possible explanation for users navigation behavior (Fu, 2012; Sundar et al., 2008). Twitter displays only the absolute number of times that a message was retweeted. However, if Twitter introduced simple icons that dichotomously express whether a tweet has been retweeted very often or not, this should lead to behavioral adaptation (i.e., social navigation) in a way that users follow the behavior of many other agents. 4. Study 1 In Study 1, we combine the possible effects of informational value and of contextual criteria, namely, agent awareness information. As mentioned earlier, prior studies have shown that if no contextual awareness information is provided, tweets that contain high informational value news factors are retweeted more often than tweets that contain low informational value news factors (e.g., Rudat, 2014; Rudat et al., 2014). This was due to the characteristic of high informational value of concerning many people and/or having the potential to affect their mind or behavior. For example, tweets like ‘‘climate change threats rice production; experts fear famines in poor countries’’ (as example for a tweet containing high informational value) was more often retweeted than tweets like this: ‘‘Spanish crown prince was spotted during bar-hopping’’ (as example for a tweet containing low informational value). Analogous to this, in Study 1, this should be the case for tweets that do not provide agent awareness information. In contrast, if tweets are marked with a star icon, then agent awareness should result in social navigation, which then should lead to a retweeting behavior that is independent from informational value. Hence, agent awareness information should moderate the influence of informational value on retweeting. Accordingly, we derived the following hypothesis: Hypothesis 1. We expect an interaction effect of Agent awareness information x Informational value. We expect that if tweets are not marked with awareness information, participants will retweet tweets containing news with high informational value more often than tweets containing news with low informational value. In contrast, we expect that if tweets are marked with awareness information, the difference between retweeted tweets containing news factors with high informational value and tweets containing news factors with low informational value will be smaller. Further, we aim to answer the following research question: Do the star icons as agent awareness information lead to higher ratings of importance of the respective tweets? This question arises when considering the potential effect of the star icons as indirect recommendation, heuristic, or norm cue. By ascribing the respective tweets high importance, the decision to follow the recommendations could be made more comfortable and consistent.

78

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

4.1. Method 4.1.1. Participants Data were collected from 64 German speaking participant volunteers (11 male, 53 female). All participants were students, recruited via a database of subscribed volunteers. For their participation in the laboratory experiment, which took 60 min, participants were paid 8 €. Alternatively, the students could receive credit for their participation if needed for course requirements. Participants’ age ranged from 18 to 37 years (M = 23.70, SD = 3.71). Participants were asked about their knowledge about Twitter and their average usage of Twitter. Both items were measured by a five point Likert-scale ranging from 1 (very little/rarely) to 5 (very good/very often). Participants indicated their average knowledge about Twitter (M = 2.05, SD = 0.98) as well as their average usage of Twitter (M = 1.36, SD = 0.76) as rather low. 4.1.2. Design A 2  2 within-subject design was used to explore the effect of informational value and agent awareness information on retweeting. For the factor informational value, we used eight news factors and subsumed them into two categories: high value category (Controversy, Negative Consequences, Relevance, and Unexpectedness) versus low value category (Aggression, Personalization, Prominence, and Proximity). For the factor agent awareness information, star icons were assigned randomly to one half of the tweets. The other half of the tweets was displayed without icon. The subset of marked tweets varied randomly for each participant. 4.1.3. Materials Material consisted of 36 tweets about a wide range of German news topics. In order to have systematically prepared and balanced material, we used self-created fictive tweets, although they often were based on real existing tweets. Because of the fictive character of the tweets, we had to use news that was credible enough without it actually having taken place. If we would have used real existing stories, the news factor Unexpectedness would not have worked, for example. For a prior set of tweets we conducted a material test in which participants rated all tweets regarding their news factors. Highly significant correlation scores (.92 > r > .59) indicated that the ascribed news factors have been well recognized. After that, several further sets of tweets have been developed and used in studies that yielded stable results (Rudat, 2014; Rudat et al., 2014). In order to ascribe the news factors to a tweet, certain trigger words were used (e.g., ‘‘brutal’’, ‘‘to insult’’, ‘‘violence’’ for Aggression; ‘‘dispute’’, ‘‘debate’’, ‘‘to put the blame on [somebody]’’ for Controversy; ‘‘loss’’, ‘‘damage’’, ‘‘threat’’ for Negative Consequences; (unknown) names for Personalization; famous names for Prominence; places in Baden-Württemberg (the region in which the study took place) for Proximity; topics such as climate, education, employment for Relevance; ‘‘unexpected’’, ‘‘despite’’, ‘‘suddenly’’ for Unexpectedness). All tweets conveyed different news factors in a different number and in different combinations. To compare the two groups of news factors, we subsumed all tweets into two non-overlapping subsets. The first set (high value set) contained all tweets that conveyed news factors with high informational value, whereas the second set (low value set) contained all tweets that conveyed news factors with low informational value. This means that each tweet contained either news factors of only high informational value or news factors of only low informational value. We prepared the material in such a way that each news factor occurred in the same number (eight times per each news factor) within all messages. Further, occurrence of news factors within one informational value category was

uncorrelated, which means that the number of occurrences of each news factor was not related to the number of occurrences of any other news factor. Examples for tweets are: ‘‘Will education suddenly become a hot topic for the federal government? (as example for a tweet containing the high informational news factors Unexpectedness and Relevance), and ‘‘Chancellor Angela Merkel will visit a museum in Stuttgart’’ (as example for a tweet containing news factors of low informational value, Prominence and Proximity). The fictive tweets were not longer than 140 characters each and looked like real tweets. The tweets supposedly were sent by (fictive) news sources. However, we omitted source icons such as logos. All tweets were presented in a simulated Twitter environment. 4.1.4. Measures As dependent variable, we measured the retweeting behavior as well as importance ratings in order to answer our research question whether star icons as social cues are rated as highly important. Retweeting: To measure retweeting information, participants had to click a button to indicate their decision to retweet the information. The decision was counted dichotomously (0 – not retweeted, 1 – retweeted). Importance ratings: We measured the importance for each tweet by asking the participants to rate each tweet on a five point Likertscale (1 – unimportant to 5 – very important). 4.1.5. Procedure We recruited participants from a database of subscribed volunteers and invited them to take part in a ‘‘Twitter – Microblogging Study’’ in a laboratory in a German research institute where they would have to read and select information given to them. All instructions and materials were presented on a computer screen. The participants were told to read the instructions and, if necessary, to ask about anything they did not understand. As the experimental manipulations in Study 1 lay solely in variations within the material, there were no different conditions for the participants, and thus, the procedure was identical. Participants were instructed about the meaning of the star icons: They were told that star icons beside the tweets mean that this respective tweet has already been retweeted very often by other Twitter users. The tweets then were presented in a random order for each participant to avoid sequence effects. We measured retweeting by the participants’ choice of retweeting a tweet or not. Participants decided by marking the checkbox of those tweets that they wished to retweet. After they had decided which tweets to retweet, all tweets were presented again. This time, participants had to rate each tweet on importance. At the end, participants were thanked and debriefed. 4.2. Results The descriptive results of the particular news factors regarding their influence on retweeting are shown in Table 2. Participants Table 2 Percentage of retweeted tweets regarding each news factor (out of eight possible ones) by each person (N = 64), alphabetically sorted. News factor

Aggression Controversy Negative Consequences Personalization Prominence Proximity Relevance Unexpectedness

Informational value

Low High High Low Low Low High High

Retweeted M

SD

.20 .38 .44 .13 .18 .17 .47 .38

.17 .23 .25 .17 .16 .16 .27 .23

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

mostly selected information for retweeting that conveyed the news factors Relevance, Negative Consequences, Unexpectedness, and Controversy. These are all news factors with high informational value. 4.2.1. Retweeting We calculated the mean retweeting scores for all tweets in all four combinations of the independent variables. Thus, we had mean retweeting scores for tweets of the high value set with star icons, tweets from the high value set without star icons, tweets from the low value set with star icons, and tweets from the low value set without star icons. All descriptive statistics regarding retweeting are presented in Table 3. To test Hypothesis 1, we performed a 2  2 factorial repeated measures analysis of variance (ANOVA), with informational value (high vs. low) and agent awareness information (present vs. absent) as independent variables and retweeting as dependent variable. According to Hypothesis 1, participants should retweet more tweets from the high value set than tweets from the low value set if tweets are not marked with agent awareness information. In contrast, if tweets are marked with agent awareness information, the difference between retweeted tweets from the high value set and tweets from the low value set should be smaller. The ANOVA revealed no interaction effect of Awareness information  Informational value, F(1, 63) = 2.39, p = .127, ns. Thus, Hypothesis 1 was not corroborated. Instead, we found a significant main effect of informational value, F(1, 63) = 120.29, p < .001, partial g2 = .656. This means that tweets from the high value set were retweeted more often than tweets from the low value set. Further, the ANOVA also revealed a main effect of agent awareness information, F(1, 63) = 9.74, p = .003, partial g2 = .134. This means that tweets that were marked with agent awareness information were retweeted more often than tweets without agent awareness information. 4.2.2. Additional analysis: Importance ratings In order to answer our research question concerning the perceived importance of tweets due to agent awareness information, we calculated mean importance rating scores for tweets with agent awareness information and tweets without agent awareness information. We performed a paired t-test with awareness information (present vs. absent) as independent variable and importance ratings as dependent variable. The analysis revealed a significant difference between the two groups, t(63) = 3.429, p = .001, d = 0.43. This means that tweets that were marked with agent awareness information (M = 3.00, SD = 0.37) were rated as more important than tweets without agent awareness information (M = 2.84, SD = 0.47). As the t-test revealed a significant difference, we conducted an additional analysis to investigate whether importance ratings mediate the relationship between awareness information and retweeting behavior, using techniques suggested by Judd, Kenny, and McClelland (2001) for testing mediation in within-subject designs. As reported above, agent awareness information affected

Table 3 Percentage of retweeted tweets. Informational value

Agent awareness information

M

SD

High

Yes No

.22 .18

.14 .11

Low

Yes No

.09 .07

.07 .09

79

importance ratings and agent awareness information also affected retweeting. Hence, the basis for testing mediation is given. Importance rating difference scores (importance ratings of tweets with awareness information minus importance ratings of tweets without awareness information) were then created. Additionally, retweeting difference scores (retweeting of tweets with awareness information minus retweeting of tweets without awareness information) were created. This retweeting difference score was then regressed on two predictors: the sum of each participants’ importance rating scores (present and absent awareness information) and the difference of each participants’ importance rating scores (Judd et al., 2001). A significant regression coefficient for the importance rating difference predictor indicated mediation of the retweeting effect by importance ratings, b = .45, p < .001. Furthermore, the estimated intercept was not found to differ from zero (B = .021, p = .76, ns), indicating complete mediation of retweeting differences by differences in importance ratings. Hence, the influence of agent awareness information on retweeting is due to the importance that participants ascribe to agent awareness information.

4.3. Discussion Study 1 investigated influencing criteria on sharing news in Twitter. Two kinds of criteria were investigated: criteria inherent in the information and contextual criteria. As information-inherent criteria, we used the concept of informational value adapted from and referring to news value theory; as context criteria, we used awareness information about agents, which is drawn from research on awareness and leads to social navigation. The agents’ behavior was marked by an icon next to the tweets, indicating that other agents have retweeted that respective tweet very often. We expected an interaction effect of awareness information and informational value in such a way that if tweets were marked with a star icon, participants should have retweeted them while disregarding the informational value of the tweets. In contrast, out of the tweets without a star icon, participants should have retweeted tweets containing high informational value news factors more often than tweets containing low informational value news factors. However, we did not find any interaction effect of awareness information and informational value. Instead, our analysis revealed two additive main effects, showing that first, high informational value led to more retweeting than low informational value and that second, star icons that indicate selection decisions of others led to more retweeting compared to no star icons. As in previous studies (e.g., Rudat et al., 2014), we found the main effect of informational value to be very strong and stable; so far, agent awareness information as contextual criteria has not been influential enough to qualify that effect. However, taking the main effect of agent awareness information together with the rating results, it appears that cues about a group of other users are indeed noticed. Moreover, although agent awareness information did not qualify the effect of informational value on retweeting, it led to a higher rating score on the importance of the tweet. It could be shown that the ratings of importance also mediated the effect of agent awareness information on retweeting. Taken together, this provides evidence for the influence of social navigation on retweeting: Aggregated traces of agents’ behavior provide orientation for other users and affect their selection decisions. Thus, for users of the Web 2.0, it might be relevant and helpful to see what other users find valuable or interesting in order to facilitate their own decision making. This finding is in line with studies that have reported on the effectiveness of awareness and social navigation cues in areas such as information seeking

80

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

(Schwind & Buder, 2012) and collaborative learning (Engelmann et al., 2009). Based on these results, one can speculate whether users’ interpretation of agent awareness rests on the perception of other agents. If agent awareness information refers to a group of people that the user identifies with, it can be assumed that the extent of social navigation should become stronger. In the next section, this assumption will be explained. 5. The influence of identification If we imagine that a new student looks for a good restaurant, she might get the information that the professors often go to restaurant A, and that other student often go to restaurant B. Which restaurant is more likely to be chosen by her? According to findings from social psychology, it should be no surprise that the students’ restaurant is much more likely to be chosen. This is due to social psychological phenomena that are known from research on social influence and social identity (Cialdini & Goldstein, 2004; Lee & Nass, 2002; Postmes, Spears, Sakhel, & de Groot, 2001; Sassenberg, 2011; Spears & Lea, 1992; Turner, 1991). According to these findings, individuals are likely to adapt their behavior to norms and rules that exist in relevant in-groups that the individuals identify with. People follow social norms in order to, for example, behave effectively, build and maintain social relationships, or in order to manage their self-concept (e.g., Cialdini & Trost, 1998; Postmes, Spears, & Lea, 2000). Therefore, recommendations from members of an identity-serving group should be more influential than recommendations that stem from group members with little or no identification. Consequently, contextual cues that refer to a relevant in-group should make the in-group membership salient and activate the corresponding norms (James & Greenberg, 1989). In this case, we argue that, although there is good reason to believe that social navigation and social recommendations per se are effective, social navigation should even be increased. Related to Twitter, a user should follow the behavior of agents who are members of a salient in-group (because he or she identifies with them) more likely than follow the behavior of agents who are not members of a salient in-group. 6. Study 2 In Study 2, our aim was to investigate whether and how the reference to a salient in-group interacts with informational value of tweets and agent awareness information regarding its influence on retweeting behavior. To do this, we conducted a second experimental laboratory study, using systematically prepared material and invited students as participants. We expected that student participants for whom the agents do not belong to a salient in-group (i.e., no reference to students) would show less social navigation than student participants for whom the agents belong to a salient in-group (i.e., reference to students). Participants without information about in-group agents should instead be more influenced by informational value compared to participants with information about in-group agents. Taken together, we expected that the influences of informational value and awareness information on retweeting behavior each will be moderated by whether the agents are members of a salient ingroup or not. Accordingly, we derived the following two hypotheses: Hypothesis 2a. The difference between tweets of high versus low informational value will be larger for student participants who were told that the star icons refer to other Twitter users compared to student participants who were told that the star icons refer to other students.

Hypothesis 2b. The difference between tweets with present versus absent agent awareness will be larger for student participants who were told that the star icons refer to other students (salient reference) compared to student participants who were told that the star icons refer to other Twitter users in general (non-salient reference).

6.1. Method 6.1.1. Participants Data were collected from 65 participant volunteers (22 male, 43 female). All participants were German speaking students, recruited via a database of subscribed volunteers. For their participation in the laboratory experiment, which took 60 min, participants were paid 8 €. Alternatively, the students could receive credit for their participation if needed for course requirements. Participants’ age ranged from 18 to 28 years (M = 21.51, SD = 2.60). Participants were asked about their average knowledge about Twitter and their average usage of Twitter. Both items were measured by a five point Likert-scale ranging from 1 (very small/rarely) to 5 (very good/very often). Participants indicated their average knowledge about Twitter (M = 2.18, SD = 1.04) as well as their average usage of Twitter (M = 1.28, SD = 0.63) as rather small. Further, participants indicated how much they identified with being students, which was also measured by a five point Likert-scale ranging from 1 (very little) to 5 (very strongly), as rather high (M = 4.00, SD = 0.94). 6.1.2. Design A 2  2  2 mixed design was used to explore the effects of informational value, agent awareness information, and in-group reference on retweeting. For the within-factor informational value, we again used the eight news factors and subsumed them into two categories: high value category (Controversy, Negative Consequences, Relevance, Unexpectedness) versus low value category (Aggression, Personalization, Prominence, Proximity). For the within-factor agent awareness information, little star icons were randomly assigned to half of the tweets. The other half of tweets was not marked with an icon. The subset of marked tweets randomly varied for each participant. For the between-factor, in-group reference, we created two experimental conditions, differing in the information about the alleged source of agent awareness information: In one group, it referred to Twitter users in general, that is, not to the salient in-group (non-salient condition, n = 32), whereas in the other group, it referred to other students, that is, to the salient in-group (salient condition, n = 33). 6.1.3. Materials The material for Study 2 was the same material as in Study 1, which consists of two non-overlapping subsets: The first set (high value set) contained all tweets that conveyed news factors with high informational value, whereas the second set (low value set) contained all tweets that conveyed news factors with low informational value. 6.1.4. Procedure and measurement We recruited student participants from a database of subscribed volunteers and invited them to take part in a ‘‘Twitter – Microblogging Study’’ in a laboratory in a German research institute where they would have to read and select information given to them. All instructions and materials were presented on a computer screen. The participants were told to read the instructions and, if necessary, to ask about anything they did not understand. The identification with the relevant in-group as a basis for manipulation was checked for all participants by asking them about their

81

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84 Table 4 Percentage of retweeted tweets regarding each news factor (out of eight possible ones) by each person (N = 65), alphabetically sorted. News factor

Aggression Controversy Negative consequences Personalization Prominence Proximity Relevance Unexpectedness

Informational value

Low High High Low Low Low High High

Retweeting M

SD

.25 .36 .43 .15 .19 .22 .44 .37

.20 .20 .23 .17 .18 .16 .25 .21

identification with being students. Then, in one condition (nonsalient condition), participants were told that the star icon refers to other Twitter users who retweeted the respective tweets very often. In contrast, in the other condition (salient condition), participants were told that the star icon refers to other students who retweeted the respective tweets very often. Otherwise the procedure was the same for all participants. The tweets were presented in a random order for each participant to avoid sequence effects. We measured retweeting by the participants’ choice of retweeting a tweet or not. To do this, they had to decide, after reading the tweets, which tweet they wanted to retweet to their followers. Participants decided by marking a checkbox adjacent to those tweets that they wished to retweet. The decision was counted dichotomously (0 – not retweeted, 1 – retweeted). At the end, participants were thanked and debriefed. 6.2. Results Before analyzing treatment effects, we checked the variables ‘‘average knowledge about Twitter’’, ‘‘average usage of Twitter’’, and ‘‘identification with being students’’ to ensure that differences were not due to pre-existing differences between the two conditions. Regarding all three items, the average knowledge about Twitter, its average usage, and identification with students, independent t-tests yielded no differences between the two conditions (knowledge: t(63) = 0.93, p = .357; usage: t(63) = 0.74, p = .464; identification: t(63) = 1.61, p = .112). For the descriptive results of the particular news factors regarding their influence on retweeting see Table 4. Participants mostly selected information for retweeting that conveyed the news factors Relevance, Negative Consequences, Unexpectedness, and Controversy. These are all news factors with high informational value. In the following, we will present the results regarding Hypotheses 2a and 2b.

6.2.1. Retweeting All descriptive statistics regarding retweeting and the hypotheses are presented in Table 5. To test the hypotheses, we performed a mixed design analysis of variance (ANOVA), with informational value (high vs. low), agent awareness information (present vs. absent), and in-group reference (non-salient vs. salient) as independent variables and retweeting as the dependent variable. According to Hypothesis 2a, participants in both the non-salient and in the salient condition should retweet tweets from the high value set more often than tweets from the low value set. However, the difference between retweeted tweets from the high and the low value set should be larger for participants in the non-salient condition than for participants in the salient condition. According to Hypothesis 2b, both participants in the non-salient and in the salient condition should retweet tweets with a star icon more often than tweets without a star icon. However, the difference between retweeted tweets with and those without a star icon should in this case be larger for participants in the salient condition than for participants in the non-salient condition. Starting with the results concerning the hypotheses, the ANOVA revealed two significant interaction effects. First, we found a significant interaction effect of informational value and in-group reference, F(1, 63) = 4.26, p = .043, partial g2 = .063. This interaction effect is in line with Hypothesis 2a and can be explained by pairwise comparisons using Bonferroni adjustment: The difference of retweeted tweets from the high and the low value set was larger for participants from the non-salient condition (F(1, 63) = 59.69, p < .001, partial g2 = .487) than for participants from the salient condition (F(1, 63) = 24.04, p < .001, partial g2 = .276). This indicates that participants from the salient condition retweeted tweets from the high value set (M = .20, SE = .02) more often than tweets from the low value set (M = .12, SE = .01). However, the difference was, as predicted, larger for participants from the non-salient condition (high value set: M = .18, SE = .02; low value set: M = .07, SE = .01). Second, and partially in line with our expectations, the ANOVA revealed a marginally significant interaction effect of awareness information and in-group reference, F(1, 63) = 3.80, p = .056, partial g2 = .057. Pairwise comparisons using Bonferroni adjustment indicated that participants in the non-salient condition retweeted tweets with (M = .13, SE = .02) and without star icons (M = .12, SE = .01) to the same extent (F(1, 63) < 1, ns), whereas participants in the salient condition retweeted tweets with a star icon (M = .18, SE = .02) more often than tweets without a star icon (M = .14, SE = .01) (F(1, 63) = 8.08, p = .006, partial g2 = .114). Thus, although the result is only marginally significant, it still shows a tendency in the direction of Hypothesis 2b.

Table 5 Percentage of retweeted tweets. Informational value

High

Low

Agent awareness information

Present M SD Absent M SD Present M SD Absent M SD

In-group reference Non-salient

Salient

Total

.19 .14

.22 .11

.20 .12

.18 .11

.17 .12

.18 .11

.06 .06

.14 .11

.10 .10

.07 .06

.11 .08

.09 .08

82

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

Further, the ANOVA also revealed three significant main effects: of informational value, F(1, 63) = 80.02, p < .001, partial g2 = .559 (high value set: M = .19, SD = .12, low value set: M = .10, SD = .09), of awareness information, F(1, 63) = 4.16, p = .046, partial g2 = .062 (tweets with a star icon: M = .15, SD = .11, tweets without a star icon: M = .13, SD = .09), and of in-group reference, F(1, 63) = 4.18, p = .045, partial g2 = .062 (salient condition: M = .16, SD = .10, non-salient condition: M = .12, SD = .09). However, all these main effects can be explained by the interaction effects described above, and are therefore qualified. We neither found an interaction effect of informational value and awareness information, F(1, 63) < 1, ns, nor a three-way interaction of informational value, awareness information, and in-group reference, F(1, 63) < 1, ns. 6.3. Discussion Study 2 investigated the moderating role of the in-group reference for the influence of informational value on retweeting and for the influence of agent awareness on retweeting. We hypothesized two interaction effects: an interaction between informational value and in-group reference (Hypothesis 2a), and an interaction between agent awareness information and in-group reference (Hypothesis 2b). Results confirmed Hypothesis 2a, and to some degree Hypothesis 2b. Student participants who were told that the agents are other Twitter users were more strongly influenced by informational value than student participants who were told that the agents are other students. In line with former findings, the study shows that informational value appears to be a powerful indicator of retweeting decisions, but the extent of this impact can be weakened by activating in-group salience. In contrast, student participants who were told that the agents are other students were more affected by agent awareness than student participants who were told that the agents are other Twitter users. If social navigation depends on whose traces the recommendations stem from, this would be in line with findings from social psychology indicating that social influence depends strongly on how valued and important the group from which the norms come is perceived (e.g., Turner, 1991). In our case, we argue that students should perceive other students as valued and important as they belong to their own social group; and it should be counterproductive for the self-esteem not to perceive one’s own group as meaningful. 7. General discussion and conclusion In the current paper, we presented two experimental studies that investigated the influence of informational value, agent awareness (Studies 1 and 2) and in-group reference (Study 2) on retweeting decisions in Twitter. Study 1 showed that high informational value leads to more retweeting than low informational value as well as that agent awareness affects retweeting decisions more compared to no agent awareness. These results fit well to results from previous studies (Rudat et al., 2014) and to results regarding the influence of social navigation and recommendations (e.g., Schwind & Buder, 2012), and the bandwagon effect (e.g., Sundar et al., 2008). However, whereas agent awareness did not moderate the influence of informational value on retweeting, in-group reference did so. Study 2 revealed that reference to a salient in-group could weaken the impact of informational value, and at the same time, could increase social navigation. An interesting difference between the results of both studies can be found by comparing the agent awareness condition of Study 1 with the agent awareness condition of Study 2 for the group of non-salience students. Whereas the former condition yielded effects of social navigation, the latter did not lead to social navigation at all. This might be

explained by the fact that in Study 2 all participants were initially asked about their identification with being students. It might be the case that this identification item created a stronger subjective contrast between oneself and a community (e.g., Lee & Nass, 2002; Spears & Lea, 1992), effectively rendering this community as an out-group in the non-salient condition. Some limitations must be considered when interpreting the results of these two studies. First, using carefully and systematically prepared material and conducting a lab studies usually means that ecological validity will be decreased. However, we would not have been able to compare retweeted and not retweeted tweets out of a well-prepared set of tweets ‘‘in the wild’’. Second, particularly in Study 1, the range of the participants’ age was quite large and thus, the sample somewhat heterogeneous. Thus might make conclusions a bit difficult. However, in Study 2, the participants’ age was more homogenous. Third, participants for the studies were mainly students who had rather low experience with Twitter. Only about 11% of them in Study 1 and about 17% in Study 2 indicated to have their own Twitter account. Therefore, further research might investigate whether this effect would also be replicable with more experienced participants. It might be speculated that experienced Twitter users have more established routines in making retweeting decision and are thus less influenced by retweeting decisions of other users. However, even the opposite might be the case: Experienced Twitter users could also be used to notice what others are doing as they are often interacting with them and thus might know that their recommendations are worth to be followed. However, we argue that in general the mechanisms we were interested in for these studies should be more or less independent of the specific setting. News factors as basis for informational value as well as principles of social navigation have turned out as being stable (e.g., Buder & Schwind, 2012; Eilders, 1997; Sundar et al., 2008). Thus, the experience of using Twitter should not be too crucial. Fourth, only one particular group was used to manipulate in-group relevance, namely, students. Future studies might investigate whether these findings are generalizable to other groups. Fifth, we did not explicitly measure the perceived importance or closeness to the group of students; we only measured the degree of identification. However, the results indicated that, at least in this case, the manipulation was successful as the group of students was recognized and had an impact on retweeting decisions. Nevertheless, future studies should extend measurements and potential influences. For example, the role of an identity-serving group (e.g., students) could be manipulated as original source of the tweets. It would be interesting to explore whether the degree of identification interacts with the source and thus influences the amount of retweeted tweets send by students. Despite these limitations, this study gave further insights into selection decisions in the Twitter context. Whether news is shared with others or not is dependent on both content criteria (informational value) and context criteria (agent awareness information). It should be noticed that the content of news, in this case, the informational value, seems to have great power for influencing selection decisions of sharing news in Twitter. However, by taking together the impact of agent awareness and of in-group reference, the results also indicate that reference to other people has the potential to shape users’ behavior, too, and thus emphasizing the importance of the social in the social Web. These insights contribute to research on news selection as many of the selection processes taking place in social media can also be described and explained with values that are derived from news value theory. Moreover, research on social media that is concerned with phenomena such as participation (e.g., Kim & Sundar, 2014), lurking (e.g., Schneider, von Krogh, & Jäger, 2013), identity and impression management (e.g., Krämer & Winter, 2008; Maireder, 2011), or knowledge management (e.g., Panahi, Watson, & Partridge, 2013)

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

might be inspired by our findings. Finally, journalists, public relations firms and advertisers could be strongly interested in our results as our findings might provide suggestions how to create and place news.

References Axsom, D., Yates, S., & Chaiken, S. (1987). Audience response as a heuristic cue in persuasion. Journal of Personality and Social Psychology, 53, 30–40. Berger, J., & Milkman, K. L. (2012). What makes online content viral? Journal of Marketing Research, 49, 192–205. Boyd, D., Golder, S., & Lotan, G. (2010). Tweet, tweet, retweet: Conversational aspects of retweeting on Twitter. Proceedings of the 43rd annual Hawaii international conference on system sciences (pp. 1–10). Buder, J. (2011). Group awareness tools for learning: Current and future directions. Computers in Human Behavior, 27, 1114–1117. Buder, J., & Bodemer, D. (2008). Supporting controversial CSCL discussions with augmented group awareness tools. International Journal of Computer-Supported Collaborative Learning, 3, 123–139. Buder, J., & Schwind, C. (2012). Learning with personalized recommender systems: A psychological view. Computers in Human Behavior, 28, 207–216. Carroll, J. M., Neale, D. C., Isenhour, P. L., Rosson, M. B., & McCrickard, D. S. (2003). Notification and awareness: Synchronizing task-oriented collaborative activity. International Journal of Human-Computer Studies, 58, 605–632. Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and conformity. Annual Review of Psychology, 55, 591–621. Cialdini, R. B., & Trost, M. R. (1998). Social influence: social norms, conformity, and compliance. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., Vol. 2, pp. 151–192). Boston, USA: McGraw-Hill. Clark, H., & Murphy, G. (1982). Audience design in meaning and reference. In J. Le Ny & W. Kintsch (Eds.), Language and comprehension (pp. 287–299). Amsterdam, Netherlands: North-Holland. Dourish, P., & Chalmers, M. (1994). Running out of space: Models of information navigation. In Proceedings of the HCI conference (pp. 23–26) . Eilders, C. (1997). Nachrichtenfaktoren und Rezeption: Eine empirische Analyse zur Auswahl und Verarbeitung politischer Information [News factors and reception: An empirical analysis of the audience’s selection and retention processes in political communication]. Opladen, Germany: Westdeutscher Verlag. Eilders, C. (2006). News factors and news decisions: Theoretical and methodological advances in Germany. Communications: European Journal of Communication Research, 31, 5–24. Eilders, C., & Wirth, W. (1999). Die Nachrichtenwertforschung auf dem Weg zum Publikum: Eine experimentelle Überprüfung des Einflusses von Nachrichtenfaktoren bei der Rezeption [News value research meets the audience: Investigating the impact of news factors on reception processes in an experimental setting]. Publizistik, 44, 35–57. Engelmann, T., Dehler, J., Bodemer, D., & Buder, J. (2009). Knowledge awareness in CSCL: A psychological perspective. Computers in Human Behavior, 25, 949–960. Fu, W. W. (2012). Selecting online videos from graphics, text, and view counts: The moderation of popularity bandwagons. Journal of Computer-Mediated Communication, 18, 46–61. Galtung, J., & Ruge, M. H. (1965). The structure of foreign news: The presentation of the Congo, Cuba and Cyprus crises in four foreign newspapers. Journal of Peace Research, 2, 64–91. Hansen, L. K., Arvidsson, A., Nielsen, F. Å., Colleoni, E., & Etter, M. (2011). Good friends, bad news: Affect and virality in Twitter. In J. J. Park, L. T. Yang, & C. Lee (Eds.). Communications in computer and information science: Vol. 185. Proceedings Part II of the 6th international conference on future information technology (pp. 34–43). Berlin, Heidelberg, Germany: Springer. Harcup, T., & O’Neill, D. (2001). What is news? Galtung and Ruge revisited. Journalism Studies, 2, 261–280. Hermida, A. (2010). From TV to Twitter: How ambient news became ambient journalism. M/C Journal, 13(2). . Höök, K., Benyon, D., & Munro, A. J. (Eds.). (2003). Designing information spaces: The social navigation approach. London, UK: Springer. James, K., & Greenberg, J. (1989). In-group salience, intergroup comparison, and individual performance and self-esteem. Personality and Social Psychology Bulletin, 15, 604–616. Janssen, J., & Bodemer, D. (2013). Coordinated computer-supported collaborative Learning: Awareness and awareness tools. Educational Psychologist, 48, 40–55. Janssen, J., Erkens, G., & Kirschner, P. A. (2011). Group awareness tools: It’s what you do with it that matters. Computers in Human Behavior, 27, 1046–1058. Java, A., Song, X., Finin, T., & Tseng, B. (2007). Why we twitter: Understanding microblogging usage and communities. In Proceedings of the 9th WebKDD and 1st SNA-KDD workshop on web mining and social network analysis (pp. 56–65). Judd, C. M., Kenny, D. A., & McClelland, G. H. (2001). Estimating and testing mediation and moderation in within-subject designs. Psychological Methods, 6, 115–134. Kim, H.-S., & Sundar, S. S. (2014). Can online buddies and bandwagon cues enhance user participation in online health communities? Computers in Human Behavior, 37, 319–333.

83

Krämer, N., & Winter, S. (2008). Impression management 2.0: The relationship of self-esteem, extraversion, self-efficacy, and self-presentation within social networking sites. Journal of Media Psychology, 20, 106–116. Kwak, H., Lee, C., Park, H., & Moon, S. (2010, April). What is Twitter, a social network or a news media? Paper presented at the International World Wide Web Conference (WWW), Raleigh, NC. Lee, C. S., & Ma, L. (2012). News sharing in social media: The effect of gratifications and prior experience. Computers in Human Behavior, 28, 331–339. Lee, E.-J., & Nass, C. (2002). Experimental tests of normative group influence and representation effects in computer-mediated communication: When interacting via computers differs from interacting with computers. Human Communication Research, 28, 349–381. Lippmann, W. (1922). Public opinion. New York, USA: Macmillan. Maier, M., Stengel, K., & Marschall, J. (2010). Nachrichtenwerttheorie. Konzepte: Ansätze der Medien- und Kommunikationswissenschaft, 2 [News value theory. Concepts: Approaches of media- and communication science, 2]. Baden-Baden, Germany: Nomos. Maireder, A. (2011): Links auf Twitter. Wie verweisen deutschsprachige Tweets auf Medieninhalte? [Links on Twitter. How do German tweets refer on media content?]. . Nadeau, R., Cloutier, E., & Guay, J.-H. (1993). New evidence about the existence of a bandwagon effect in the opinion formation process. International Political Science Review, 14, 203–213. O’Reilly, T. (2005, September 30). What is Web 2.0: Design patterns and business models for the next generation of software. . Panahi, S., Watson, J., & Partridge, H. (2013). Towards tacit knowledge sharing over social web tools. Journal of Knowledge Management, 17, 379–397. Papacharissi, Z., & de Fatima Oliveira, M. (2012). Affective news and networked publics: The rhythms of news storytelling on #Egypt. Journal of Communication, 62, 266–282. Phielix, C., Prins, F. J., Kirschner, P. A., Erkens, G., & Jaspers, J. (2011). Group awareness of social and cognitive performance in a CSCL environment: Effects of a peer feedback and reflection tool. Computers in Human Behavior, 27, 1087–1102. Postmes, T., Spears, R., & Lea, M. (2000). The formation of group norms in computermediated communication. Human Communication Research, 26, 341–371. Postmes, T., Spears, R., Sakhel, K., & de Groot, D. (2001). Social influence in computer-mediated communication: The effects of anonymity on group behavior. Personality and Social Psychology Bulletin, 27, 1243–1254. Rosengren, K. E. (1974). International news: Methods, data, and theory. Journal of Peace Research, 11, 145–156. Rössler, P., Bomhoff, J., Haschke, J. F., Kersten, J., & Müller, R. (2011). Selection and impact of press photography: An empirical study on the basis of news factors. Communications, 36, 415–439. Rudat, A. (2014). Twitter breaks the news: Influencing factors on sharing news in Twitter (Doctoral dissertation). Retrieved from https://publikationen.unituebingen.de/xmlui/handle/10900/50718. Rudat, A., Buder, J., & Hesse, F. W. (2014). Audience design in Twitter: Retweeting behavior between informational value and followers’ interests. Computers in Human Behavior, 35, 132–139. Ruhrmann, G., & Göbbel, R. (2007). Veränderung der Nachrichtenfaktoren und die Auswirkungen auf die journalistische Praxis in Deutschland [Change of news factors and effects on journalistic routines in Germany]. Wiesbaden, Germany: Netzwerk Recherche. Sande, Ø. (1971). The perception of foreign news. Journal of Peace Research, 8, 221–237. Sassenberg, K. (2011). An interactional approach to social influence in computermediated communication. In Z. Birchmeier, B. Dietz-Uhler, & G. Stasser (Eds.), Strategic uses of social technology: An interactive perspective of social psychology (pp. 63–83). New York: Cambridge University Press. Schneider, A., von Krogh, G., & Jäger, P. (2013). ‘‘What’s coming next?’’ Epistemic curiosity and lurking behavior in online communities. Computers in Human Behavior, 29, 293–303. Schreiber, M., & Engelmann, T. (2010). Knowledge and information awareness for initiating transactive memory system processes of computer-supported collaborating ad hoc groups. Computers in Human Behavior, 26, 1701–1709. Schwind, C., & Buder, J. (2012). Reducing confirmation bias and evaluation bias: When are preference-inconsistent recommendations effective – And when not? Computers in Human Behavior, 28, 2280–2290. Shah, A. K., & Oppenheimer, D. M. (2008). Heuristics made easy: An effort-reduction framework. Psychological Bulletin, 34, 207–222. Shoemaker, P. J. (2006). News and newsworthiness: A commentary. Communications, 31, 105–111. Spears, R., & Lea, M. (1992). Social influence and the influence of the ‘social’ in computer-mediated communication. In M. Lea (Ed.), Contexts of computermediated communication (pp. 30–65). Hertfordshire, England: Harvester Wheatsheaf. Staab, J. F. (1990). The role of news factors in news selection: A theoretical reconsideration. European Journal of Communication, 5, 423–443. Sundar, S. S., Oeldorf-Hirsch, A., & Xu, Q. (2008). The bandwagon effect of collaborative filtering technology. In Proceedings of the 26th annual CHI conference on human factors in computing systems (pp. 3453–3458). Sundar, S. S., Xu, Q., & Oeldorf-Hirsch, A. (2009). Authority vs. peer: How interface cues influence users. In Proceedings of the 27th annual CHI conference on human factors in computing systems (pp. 4231–4236).

84

A. Rudat, J. Buder / Computers in Human Behavior 46 (2015) 75–84

Suh, B., Hong, L., Pirolli, P., & Chi, E. H. (2010). Want to be retweeted? Large scale analytics on factors impacting retweet in Twitter network. Proceedings of the 2010 IEEE 2nd International Conference on Social Computing, 177–184. Turner, J. C. (1991). Social influence. Belmont, USA: Thomson Brooks/Cole Publishing Co. Westerman, D., Spence, P. R., & Van Der Heide, B. (2012). A social network as information: The effect of system generated reports of connectedness on credibility on Twitter. Computers in Human Behavior, 28, 199–206.

Winter, S., Krämer, N. C., Appel, J., & Schielke, K. (2010). Information selection in the blogosphere: The effect of expertise, community rating, and age. In S. Ohlsson & R. Catrambone (Eds.), Proceedings of the 32nd annual conference of the Cognitive Science Society (pp. 802–807). Austin, USA: Cognitive Science Society. Zarella, D. (2009). The science of retweets. .