Transactions on Human-Computer Interaction AIS Transactions on Human-Computer Interaction THCI Original Research Audience Gatekeeping in the Twitter Service: An Investigation of Tweets about the 2009 Gaza Conflict K. Hazel Kwon Arizona State University khkwon@asu.edu Onook Oh University of Nebraska onookoh@gmail.com Manish Agrawal Unversity of South Florida magrawal@coba.usf.edu H. Raghav Rao SUNY Buffalo mgmtrao@gmail.com Abstract Twitter is a social news service in which information is selected and distributed by individual members of the tweet audience. While communication literature has studied traditional news media and the propagation of information, to our knowledge there have been no studies of the new social media and their impacts on the propagation of news during extreme event situations. This exploration attempts to build an understanding of how preexisting hyperlink structures on the Web and different types of information channels affect Twitter audiences’ information selection. The study analyzes the concentration of user-selected information sources in Twitter about the 2009 Israel-Gaza conflict. There are three findings. First, a statistical test of a power-law structure revealed that, while a wide range of information was selected and redistributed by Twitter users, the aggregation of these selections over-represented a small number of prominent websites. Second, binomial regression analyses showed that Twitter user selections were not constituted randomly but were affected by the number of hyperlinks received and the types of information channels. Third, temporal analyses revealed that sources via social media channels were more prominently selected especially in the later stages of the news information lifespan. Keywords: Twitter service, social media, audience gatekeeping, online journalism, information concentration, Israel-Gaza conflict Ping Zhang was the accepting Senior Editor. This article was submitted on 3/4/2012 and accepted on 9/27/2012. It was with the authors 80 days for 2 revisions. An early version of the paper was presented at ICIS 2011 in Shanghai and was fast tracked to THCI. Kwon, K. H., O. Oh, M. Agrawal, and H. R. Rao (2012) “Audience Gatekeeping in the Twitter Service: An Investigation of Tweets about the 2009 Gaza Conflict,” AIS Transactions on Human-Computer Interaction (4) 4, pp.212-229. Volume 4 ■ Issue 4 ■ December 2012 Kwon et al. Audience Gatekeeping in the Twitter Service INTRODUCTION The rise of social media in general and Twitter in particular has allowed the online public to actively participate in processes of news creation and circulation. Launched in 2006, Twitter is an online information sharing service accessible through various digital and mobile devices (Mills et al., 2009; Li and Lao, 2010). Users can easily exchange short messages called “tweets” that are limited to a maximum of 140 characters. Twitter is composed of constellations of users’ personal networks in which the users play a dual role of both audience and broadcaster. In the audience role, users receive information from within their Twitter personal network as well as from external sources outside Twitter. As broadcasters, they create or relay information to their own follower networks. Consequently, user-originated tweets include not just self-produced texts but also works of others that are 1 redistributed either by manual posting or by using “retweet” (or RT) or sharing “plug-ins” embedded in the designated websites. The current study explores what kinds of web sources Twitter audiences adopt to understand a news event. A news event is narrated differently according not only to the storyteller’s preferences but also to the characteristics of the media channel that conveys the story. Therefore, a Twitter audience’s understanding of an event is partly affected by which delivery channels are widely adopted. This inquiry is important especially given the elevated status of Twitter audiences in constructing and disseminating public information. The influence of social media users, including Twitter users, upon the conventional methods of news creation and dissemination is also acknowledged by traditional mass media sectors. For instance, according to Tony Maddox, the executive vice president of CNN International, CNN has been strategically “covering reports on social media networks such as Twitter and Facebook” as a means to provide audiences with a “sense of involvement and putting them right in the moment” (Shin and Yim, 2011, n.p.). While recent Twitter studies have primarily looked at how the textual messages produced within Twitter contribute to self-promotion and audience management (e.g., Marwick and Boyd, 2011) and collaborative works (e.g., Oh et al., 2010), few studies have explored the dynamics of how already available online news items are re-circulated in Twitter. User engagement in re-circulation of existing web content is a widespread practice; Smith and Boyle (2012) reported that a majority of mobile phone users are constantly connected online to watch, verify, comment, and disseminate information. Shoemaker and Vos (2009) called the practice “audience gatekeeping,” in which users “pass along already available news items and comment on them” based on the user’s own set of criteria about the newsworthiness (p. 113). Twitter is one of the popular channels through which audiences engage in such a gatekeeping process. Understanding Twitter as an audience gatekeeping channel helps to clarify the status of an audience in the contemporary social media environment. Journalism scholars in the past conceptualized the word-of-mouth among audiences as small-scale interpersonal conversations subsequent to the reception of mass-mediated information products (Southwell and Yzer, 2007). The role of audience communication, therefore, used to be considered to have little influence on determining newsworthiness. In a social media environment, however, user-to-user communication is acknowledged not just as a consequential behavior of news reception but as an additional step in a chain of news gatekeeping. According to a survey, the increased role of the online public is so evident that more than one third of users have contributed to the process of news creation and dissemination via social media sites (Purcell et al., 2010). The contributions of this paper are twofold: First, theoretically, we adapt the theory of audience gatekeeping (Shoemaker and Vos, 2009; Shoemaker et al., 2011) to envisage personal Twitter users as agents who re-distribute information items based on their own criteria. Second, among possible elements that comprise an audience’s selection criteria, we study two media centric factors that might affect selection routines and analyze the subsequent configuration of media concentration pattern. One of the two factors is the existing web structure, particularly the hyperlink networks. The other is the type of online channel through which information or news is acquired. To explore the effects of these two factors, we analyze Twitter use in the context of one of the most persistent international news stories, that of the relationship between Israel and Palestine. Specifically, we focus on the Gaza conflict of 2009 and examine a short-window event that covers the time period of the active conflict. We chose the Gaza conflict of 2009 for three reasons: One, this conflict is known as one of the signposts regarding the evidence of utility of micro-blogging as an information delivery system during extreme incidents (Israel, 2009). Two, the relationship between Israel and Palestine is a significant international political issue which various information providers, not to mention mainstream news organizations, find newsworthy. Three, although the conflict itself begins as an instantaneous event, the complicated geopolitics in this region lead the related news coverage to take “thematic news frames,” which focus not just on snapshots of concrete instances but general conditions and longterm implications of the issue (Iyengar, 1994, p.14). Thematic news frames involve multifaceted and recurrent elements that are likely to be covered for a longer period of time than any instantaneous or episodic news such as a natural disaster, car accident, or violence crime. Given the thematic nature of the topic, the Gaza conflict is a good fit to investigate the over-time pattern of audience news selection. AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 213 Kwon et al. Audience Gatekeeping in the Twitter Service TWITTER AND AUDIENCE GATEKEEPING Participatory social media embrace amateur online users as information co-creators by harnessing the users’ selfauthoring, sharing, and collective intelligence (O’Reilly, 2005). Bruns (2007) describes social media users with his own term “producers,” i.e., users who are simultaneously consumers and producers. In this view, producers are the online public who are empowered by web 2.0 technologies that enable practices of searching, linking, authoring, tagging, recommending, disseminating, and updating in an easier way and at minimal cost (Brynjolfsson and McAfee, 2007). Online participation in the process of news diffusion is generally understood in two ways. First, users play the role of alternative news producers. A representative example is the bottom-up journalism at sites such as Moveon.org or Ohmynews.com, in which citizens seek an event to construct their own reporting. Often, the narratives in citizen journalism reveal discrepancies between ordinary citizens’ versions and elite media versions of events (Kwon and Moon, 2009), triggering public allegations of bias in corporate media (Chang, 2005). Although one of the important tasks of news media has been to set up the criteria of news worthiness and to frame newsworthy issues, the legitimacy of such criteria is no longer taken for granted in the decentralized online environment which “negates the role of a central news gatekeeper” (Meraz, 2009, p.684). Indeed, in recent years several extraordinary political and social issues have occurred that have been exposed to the public primarily through social reporting. For example, information about the recent multiple protests in Arabic countries was transmitted largely via wikileaks.ch, ushahidi.com, Twitter, and Facebook. Cases such as this reflect the fact that traditional mainstream media have been experiencing weakened power to control information flow, while the citizens who used to be conceived as mere recipients have emerged as alternative information producers. Another way online participation in the process of news diffusion may be understood is through the practice of “audience gatekeeping” (Shoemaker et al., 2011). As an audience gatekeeper, the user’s role is not to create news items but to filter and deliver existing content to other audiences. As defined by McQuail (1994), gatekeeping traditionally refers to “the process by which selections are made in media work, especially decisions regarding whether or not to admit a particular news story to pass through the ‘gates’ of a news medium” (p. 213). Various gatekeeping theories have commonly assumed that news is differentiated from raw information because news is storytelling constructed by professional journalists who reorganize pieces of information selected from different sources (Shoemaker and Vos, 2009). Therefore, gatekeeping begins when a journalist becomes aware of the existence of an event (Halloran et al., 1970). Based on that awareness, professional journalists gather raw information from their own experiences, private enquiries, government resources, newswire or other media organizations, etc. Traditionally, most of the raw sources are not likely to reach audiences until they are reconstructed as news by news organizations. Accordingly, understanding how raw information is processed into a news story has been a critical inquiry that assesses the quality of social representation in media. Research on the daily editorial process in a newsroom suggests that multilevel factors should be considered in their decision making, ranging from the editorial staff ’ s personal preferences, media or channel characteristics, institutional and economic infrastructures, and ideological orientation of the community of which the news organization is a part (Clayman and Reisner, 1998; Shoemaker and Reese, 1996). The traditional gatekeeping models exclusively focus on the role of professional media workers as gatekeepers and the factors that affect their decision-making. Therefore, audiences have been excluded from the theoretical conceptualization. Barzilai-Nahon (2008) insightfully pointed out that the absence of vocabulary that refers to the message recipients subjected to the gatekeeping effect reflects the relative negligence toward this entity in the traditional gatekeeping literature. She calls the recipients of the processed information the “gated” (p.1496). In her view, although the gated are the message receivers, they are not the last stop. Instead, they do intervene in the gatekeeping process to varied degrees depending on their level of “political power,” ability of “information production,” “relationship” with traditional gatekeepers, and ability to find and choose “alternatives” as substitutes for elite news content (pp.1500-1501). Shoemaker and Vos (2009) also pointed out the unassigned role of audiences in traditional models. They call for one more mechanism: “audience’s gatekeeping, referring to audience members providing information to each other about their favored news items” (Shoemaker et al., 2011, p.61). Audience gatekeeping emerges as an important process particularly on the Internet where user sharing has an even greater impact on determining the importance of news agendas than conventional forms of interpersonal communication. The audience gets involved in gatekeeping by emailing news items that they select, or by sharing them through social media channels such as digg.com, reddit.com, newsvine.com, and twitter.com (Shoemaker and Vos, 2009). According to Goode (2009), audience gatekeeping is a form of “metajournalism” with the primary purpose of expanding the circulation of already existing information by leveraging aggregation algorithms of web 2.0 and public participation (p. 1290). AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 214 Kwon et al. Audience Gatekeeping in the Twitter Service Once audience gatekeeping is included in the bigger picture of news production and distribution, content made by media professionals should not be regarded simply as a finalized product anymore. Rather, professional news sites becomes one option for users to select along with other channels that used to be raw information providers accessible only to media professionals in the past (Goode, 2009; Southwell and Yzer, 2007). In the contemporary online environment, therefore, the audience is exposed to various information options such as news articles processed by media professionals, social reporting by citizen journalists, and raw sources directly accessible from the respective online databases. Audience gatekeeping is similar to what Bruns (2005) called “gatewatching” which refers to the practice of information gathering rather than creating, analogous to “the specialist librarian, who constantly surveys what information becomes available in a variety of media and serves as a guide to the most relevant sources when approached by information seekers” (p.7). In Twitter, audience gatekeeping occurs by posting the URL (i.e., hyperlinking) of the selected external content, or by re-tweeting within Twitter. Dimitrova, et al. (2003) argued that hyperlinking to other webpages is a part of gatekeeping that influences readers’ information choice. While their study is based solely on professional journalism sites, their discussion is expandable to social news sites in which users’ hyperlinking influences other users’ information choices. During audience gatekeeping in Twitter, professional news contents compete for user attention with other online sources, some of which used to be merely raw materials available to the public only after being processed by media professionals. Also, a growing amount of user-generated content is added to the pool of selections. Once selected, the content is often reconstructed by the addition of the user’s own comments. The reconstructed content is tweeted to other users who were yet unspoken, with the potential of reproduction within the Twitter community via re-tweeting (Figure 1). Figure 1: Twitter Audience Gatekeeping RESEARCH BACKGROUND AND HYPOTHESES Shoemaker et al. (2011) studied the popularity of news events among audiences and concluded that news items’ popularity, represented by rank order based on the frequency at which audiences share each item, is a collective indicator of newsworthiness according to audiences. While these researchers focused on each news event as the unit of analysis, another important line of inquiry into the dynamic of audience gatekeeping is the popularity of information channels, given that storytelling of the same issue can differ depending on the characteristics of the medium that delivers the issue. Despite vastly fragmented channels available online, the majority of the audience has a penchant for accessing content delivered through broadly appealing or already popular channels rather than selecting channels based on specific topical or content interest (Webster and Ksiazek, 2012). In other words, audiences’ criteria for channel AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 215 Kwon et al. Audience Gatekeeping in the Twitter Service selection are not free from media-centric factors. While various factors can potentially be considered to define users’ criteria, few studies have investigated those factors and how they impact the audience gatekeeping process. As an early attempt in this line of research, this paper is specifically focused on two factors of interest: the type of media channel and web hyperlink structure. The Impact of Channel Types on Twitter Gatekeeping A potential factor behind audience gatekeeping is the nature of channels through which the audience receives information. For example, users may be more likely to select major news organizational websites than other alternative providers even though all of them are easily accessible online. According to Webster and Lin’s (2002) study on audiences’ cable channel selections, the majority of audiences tend to insist on consuming familiar mainstream channels rather than taking a risk by adopting novel alternatives due to the passivity inherent in audience behaviors and the presumption that the high production budgets of mainstream media are the “manifestation of information quality” (Napoli, 2008, p. 58). Yim’s (2003) study on cable channels also found that “item diversity” does not necessarily lead to “exposure diversity” (p.126). In contrast, the abundance of choices actually results in an even greater audience concentration on a handful of channels, because an individual’s channel repertoire – a subset of items that an individual regularly visits – is not randomly constructed (Yim, 2003). For example, the major broadcasting network channels are almost always included in an individual’s cable channel repertoire. However, Twitter can facilitate searching through obscure information sources by amortizing search costs among a large number of users. The reputation of a Twitter user is likely to be increased by the user’s ability to highlight interesting but previously unknown information. This therefore motivates users to search for interesting information in obscure channels and bring it to the attention of the wider audience. We therefore posit a research question: • RQ1: Will Twitter users’ channel selection be concentrated more on the traditional media-based websites than other alternative types of websites? The Impact of Web Hyperlink Structure on Twitter Gatekeeping On an individual level, hyperlinking is simply a referencing practice that “automatically brings the user to a particular point in a cited work” (Halavais, 2008, p.39). On a collective level however, the hyperlink network is a form of media measurement and the most fundamental mechanism of online gatekeeping (Webster and Ksiazek, 2012; Zittrain, 2006). Sundar and Nass (2001) pointed out that collective user behavior is an important component in the evaluation of the credibility of an online source, given that it is partly “responsible for the content floating around in any given media vehicle” (p. 59). A majority of the online audience agrees that the vastness of online content is an overwhelming experience (Purcell et al., 2010). Subsequently, audiences often inevitably rely on search engine results to select a manageable amount of content from the clutter. Considering that many search engines de facto operate under the algorithms ruled by hyperlink traffic, for instance Google’s PageRank system (Finkelstein, 2008) which gives higher ranking to pages with a higher count of inlinks, the link structure configured in the Web is likely to affect Twitter users’ information selection. Therefore, we hypothesize that: • H1: The number of hyperlinks received by a certain web channel will be positively associated with the popularity of the channel among Twitter audiences. The Interaction Effect of Hyperlinks and Channel Types on Twitter Gatekeeping Hyperlink structure and channel type are likely to interact with each other to influence audience gatekeeping in Twitter. For example, while mass media-based websites are likely to be culturally familiar channels to many users, only a few of them may be preferred to other types. Users might prefer certain mass media channels when they also receive enough link traffic to be ranked high from the search results. Likewise, even though the search results position a certain website high up in the order of presentation, users may not consider it the most relevant if they have never heard of the source site previously. In these cases, there is an interaction between the content provider types and the hyperlinks. Therefore, we hypothesize that: • H2: The interaction between the degree of hyperlinks received by a channel and the channel type will be associated with the popularity of the channel among Twitter users. Temporal Differences in Audience Gatekeeping Timeliness is an important news value considered in the process of gatekeeping. Even when treating the same event, the editorial board evaluates news sources differently over a period of time because the attributes to be highlighted AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 216 Kwon et al. Audience Gatekeeping in the Twitter Service vary depending on the phase of the event’s progress (Chyi and McCombs, 2004). In a similar vein, audience gatekeepers are likely to find different information channels interesting during different phases of the development of the event. For instance, audiences might consider local information avenues to be useful in the early phase when instantaneous updates of specific events could be the main news agenda. On the other hand, channels that offer indepth reports and experts’ interpretations can become more meaningful as the event is developed into an investigative theme. In other words, audiences’ channel selection can change over time. We suggest that the temporal aspect will result in differences in Twitter users’ choice of information channels. • H3: Popular information channels among Twitter audiences will differ depending on the developmental stage of the news event. The Macro-Pattern of Channel Popularity among Twitter Audiences As mentioned earlier, our investigation is on a macro scope. We are questioning what kind of selection pattern will be configured as a collective result of audience gatekeeping. One possibility is a power-law distribution, which is a commonly found pattern on the web (Barabasi et al., 2000). A power-law distribution is characterized as being composed of a few highly popular selections and a large number of peripherals. This distribution pattern is also called the ‘preferential attachment tendency,’ indicating that people make choices preferably based on preexisting popularity. Subsequently, the pattern follows a “rich-get-richer mechanism,” resulting in severe inequality in attention distribution (Easly and Kleinberg, 2010, p.566). The power-law distribution pattern has been found in many subsets of online communities. In her study on the blogosphere, Meraz (2009) found that despite the technological affordance of decentralization, the emergent structure shows inequality with the celebrity status of so-called “A-list” bloggers and the marginalization of everyone else. Hindman (2008) also noted that the political use of the Internet shows uneven traffic among the relevant websites. We ask whether audience gatekeeping in Twitter, as a sub-practice in Websphere, will reveal a similar pattern: • RQ2: Will Twitter users’ selection of information channels collectively produce a power-law distribution? ISRAEL-GAZA CONFLICT AND TWITTER This study analyzes Twitter data pertaining to the Israel-Gaza conflict over a short event window (December 27th, th 2008 to January 18 , 2009). The topic was selected because the Israel-Palestine discord has been one of the most long-lasting and recurring international news topics for decades. Therefore, even though one particular disaster may be breakout news, the issue is situated in a complex geo-political conflict, making the temporal analysis meaningful. The hostile relations between Israel and Palestine have been widely reported in online media as well. According to Kang and Choi (1999), Israel was ranked second among the investigated 45 countries in terms of international news coverage via online news groups, with Palestine and Israel frequently mentioned concurrently in news articles. This topic is also relevant since research on Twitter use in the context of manmade extreme events is sparse (Oh et al., 2011; Prentice et al., 2011), although prior research has studied tweets in the context of natural extreme events (Mills et al., 2009 such as the Haiti earthquake (Oh et al., 2010) and the China earthquake (Li and Rao, 2010). The Israel-Gaza conflict in 2009 started with Israel’s major military attack against Hamas in the Gaza Strip with the claim of suppressing rocket fire into the southern Israel territory from the Gaza Strip (Zanotti et al., 2009). Due to the sensitive political nature of those areas, the Israel-Gaza conflict has drawn global attention from both mainstream and social media. Despite a Supreme Court ruling, Israel coordinated a media campaign in response to this attack to prevent foreign journalists from accessing the Gaza strip area 2, and Israeli soldiers were prohibited from carrying mobile phones which could unintentionally leak embarrassing information to the world (Ward, 2009). Ironically, such media control forced outsiders to rely on the public driven social media (i.e., blogs, Twitter, Facebook, YouTube, etc.) even more (Ward, 2009). Given the diversity of online information sources produced during the course of the event, the Israel-Gaza conflict can be a good case in which to study the online public’s source selection behavior as a means to produce, share, and consume related news. METHODS Data Collection and Cleaning Using Twitter’s search engine (http://search.Twitter.com), we first collected a total of 6,839 Twitter messages that were sent during the Israel-Gaza conflict. To collect comprehensive but relevant sample data, we monitored and searched for news about the conflict using Google’s search engine. Through the search, we identified that various AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 217 Audience Gatekeeping in the Twitter Service Kwon et al. news media used different terms to describe the conflict. We used those terms to perform real-time searches via Twitter’s search engine. The final search keywords that we used for data collection and the resulting number of tweets for each term are as follows: Gaza Air Strike (98), Gaza Attack (840), Gaza Ceasefire (1146), Gaza Clash (62), Gaza Conflict (954), Gaza Emergency (164), Gaza Massacre (480), Gaza Violence (648), and Gaza War (2447). We intentionally did not contain the country names ‘Israel’ and ‘Palestine’ in any search keyword, and we excluded any local language such as Hebrew or Arabic to prevent the dataset from any systematic bias that could result in over-representation of either a pro-Israel or pro-Palestine view. Considering that our primary interest is the impact of audience gatekeeping on the popularity of information channels, the unit of analysis is discrete media channels. Therefore, we first extracted tweets that included references to external online websites, resulting in a total of 2,297 tweets retained. External websites are often posted to tweets in 3 tiny URL form. For this study, each tiny URL was translated into the original full URL. Then, the URL was cleaned, leaving only the registered and the top one or two-level domain names (http://www.zzz.zzz/ or http://www.zzz.zzz.zzz/), deleting characters that followed the first slash (/). There were 115 tweets that included unidentifiable URLs, and we removed them from the sample. Among these tweets, we additionally separated the tweets posted by personal Twitter users. First, we removed unidentifiable profile accounts. We did not consider accounts as “personal” if they represented any organizational or media entity or a computer-automated bot. However, to discuss the distinctive aspect of ordinary users’ gatekeeping, these non-personal accounts were additionally analyzed and compared to the personal user-based data. As a result, 860 tweets were considered as personal users’ tweets. To conduct temporal analyses, the data set was split into two halves. Given that a large number of news stories were circulated during the few days immediately following the incident, a time-based split would have been imbalanced, making the temporal comparison hard to interpret. Therefore, we divided the data into the two chronologically ordered groups which contained an equivalent amount of tweets. The first half was considered to be the early stage of news dissemination (from December 27th, 2008 to January 6th, 2009) and the rest as the later stage (from January 7th, 2009 th to January 18 , 2009). Variables The unit of analysis is each information channel, or selected content website. Therefore, we created a list of websites based on their domain names from the tweets collected. A total of 256 unique websites were identified and each variable associated with the websites is described below. Dependent variable The dependent variable in this study is channel popularity, indicated as the frequency of each website tweeted. For example, if a website is tweeted 40 times, the website is understood to be a more popular channel among Twitter users than another website which is tweeted only 20 times. As mentioned above, each website was identified based on its domain name taken from the full URL of the tweeted web page. Independent variable There are two independent variables in this study. 1) Hyperlinks: To measure this variable, we retrieved the numeric data indicating the number of hyperlinks that a particular website received from other websites from Alexa.com, a web information company. This data contains global traffic. Given the severe skewness and heterogeneity of variances, we log-transformed the data to fit into the analysis (before transformation, M = 35609.53, SD = 1.16330E5, and after transformation, M = 7.69, SD = 2.75) 2) The types of channels tweeted 4: In order to determine types of tweeted websites, we analyzed their formats and self-descriptions and categorized them heuristically into five types. Two graduate students were trained to code data based on the five categories. The inter-coder reliability was satisfactory, Cohen’s Kappa k = .890, p < .001. Table 1 provides brief descriptions of and exemplary websites in each category. (a) Traditional News Organizations (both broadcasting and print): This category subsumes conventional mass mediaaffiliated websites, including newswire, broadcasting, and print, whose headquarters are located offline and run by media corporations or public broadcasting networks. Print media was required to have an offline edition. The websites under this category target mass audiences by presenting a broad range of topics such as politics, culture, society, travel, technology, entertainment etc. (b) Commercial Social Media: Certain formats were identified by domain names whose websites offered space for user-generated content, such as social networking, visual/audio updating, blogging, micro-blogging, and/or content AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 218 Kwon et al. Audience Gatekeeping in the Twitter Service aggregate services. Examples are youtube.com, facebook.com, yahoo.com, digg.com, reddit.com etc. Twitter also falls into this category. (c) Online Journalism: Chang (2005) identifies three different types of online journalism: first, traditional media that use electronic means to transmit messages; second, online newspapers that adhere to journalistic writing and formatting, but do not have an offline edition (which Chang calls “genuine online journalism”); and third, informal online communities where ordinary users participate in information exchanges such as newsgroup and bulletin boards (p. 926). In our study, the first type is categorized as “traditional media” and the third type overlaps with other categories. Therefore, the category “online journalism” as we have conceptualized it only includes the second type that Chang defined. Most websites in this category only had online editions. However, in a few occasions, websites did not clearly specify whether they were exclusively online or whether they had offline editions. In these cases we included websites in this third category if the content revealed a sub-cultural audience taste despite the journalistic writing style. (d) Personal Reports: This type of channel has its own domain name and is run by either an individual or a small number of people. These websites are informal and do not show formal organizational characteristics. (e) Other Organizational/Institutional/Community Websites: This type includes government, non-government, education, research, advocacy organization, and community websites. Some advocacy websites play a dual function as citizen journalism and movement organizers. In this case, the role presented on the front page was used to determine which category should be assigned to the website. If a primarily journalistic writing style was presented on the front page, it was coded as online journalism. If mobilization efforts such as petition, protest information, or donation were more prominent on the front page, the website was categorized as “other.” Table 1: Descriptions of Channel Categories Channel Type Traditional News Organizations • • • • Description Examples Newswire, broadcasting, and print mass media. Must have offline presence (e.g. offline-based headquarters, offline edition of newspapers). Should target mass audiences and include a broad range of topics on the websites. cnn.com aljazeera.com guardian.co.uk Offers a space for users to easily create and share contents. Sometimes have an automatic function that aggregates user-generated contents or external information and/or facilitates sharing processes. youtube.com reddit.com facebook.com Commercial Social Media • Online journalism • • Journalistic writing style yet no offline edition. Have independent domain names. alternet.org huffingtonpost.com allvoices.com • • Informal websites run by either an individual or a small number of people. Have independent domain name. Do not have formal organizational structure. polizero.com andycarvin.com buzzsuggest.com Any organizational / institutional / community websites not categorized above. For example, governmental, corporate, educational, research, advocacy organizational websites. gazatalk.com un.org arabmediasociety.com • Personal Reports Other Organizational/ Institutional Websites • • RESULTS Data Description The frequency analysis revealed that 256 unique websites were identified from personal users’ tweets. Interestingly, fewer than 30% of the total tweets referred to traditional news organization websites. Specifically, 76 (29.7%) out of 256 were traditional media, 40 (15.6%) were social media sites, 43 (16.8%) online journalism, 39 (15.2%) were personal content providers, and 58 (22.7%) were other organizational websites. On the other hand, we found that AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 219 Kwon et al. Audience Gatekeeping in the Twitter Service when examining all tweets, traditional channels accounted for a much larger proportion: 43.18% of the total selections. This suggests that whereas mass media entities aggressively use Twitter as another broadcasting venue, personal users’ selection process is not necessarily mass-media dominant but more inclusive of diverse types of sources. When the two time periods of news dissemination defined above were compared, “social media” had the highest mean frequency in both periods, and the frequency sharply increased as time passed: in the early stage, the mean frequency of social media was 3.43, while it was 5.75 in the later stage: t(49) = 20.31, p < .001. Also, the mean frequency increased for “organizational, institutional, and community websites” from 1.88 to 2.45 as the news evolved: t(67) = 24.81, p < .001. Overall, the most frequently tweeted websites included: youtube.com (41 times), blogspot.com (40 times), bbc.com (37 times), and friendfeed.com (35 times), three of which were categorized as “social media sites.” Table 2: Mean Frequencies and Hyperlink counts for Each Content Type in Different Time Stages Content Types Time 1 Time 2 Mean FRQ Mean Hyperlink N Mean FRQ Mean Hyperlink N Traditional Media 2.92 9.34 48 2.64 9.27 55 Commercial Social Media 3.43 10.74 28 5.75 9.30 22 Online Journalism 1.84 8.12 25 1.43 8.42 23 Personal 1.05 5.12 21 1.10 5.88 22 Org/Inst /Community 1.88 6.66 33 2.47 8.21 35 Total 2.36 8.27 155 2.45 8.14 157 The majority of providers, (68.8%), were tweeted only once. Appendix A lists the top 20 most tweeted websites along with hyperlink scores and channel types. The Effects of Hyperlinks and Channel types on Channel Popularity To examine the effects of hyperlinks and channel types, we ran negative binomial regression models based on the two time periods. When a dependent variable is a count variable, as presented in our data, ordinary least squares regression cannot be performed due to the violation of normality in residuals. Instead, Poisson regression modeling is a common statistical approach. However, Poisson regression requires strict adherence to the assumption of dispersion by which the expected mean value should be approximately equal to the observed variance. To quantify the equality of variance to the expected mean, it should be approximately equal to one when the residual deviation is divided by the degrees of freedom. If the result is greater than one, the data fail to fit the Poisson distribution assumption due to over-dispersion (Berk and MacDonald, 2008). Our dataset was over-dispersed; the ratio of residual deviance (767.76) to degrees of freedom (238) was 3.23. In this case, the use of a negative binomial regression model is recommended. Therefore, a negative binomial regression for the two time periods was conducted, with hyperlinks and types as the independent variables and channel popularity as the dependent variable. The variable “channel types” is a categorical variable, requiring a reference level for comparison. We chose “traditional news organization” and “social media” as references for interpretations, because (1) audiences’ familiarity might lead “traditional news organization” to be tweeted more than other types and (2) “social media” showed the highest mean frequency among all. The 2 omnibus tests showed a good fit for all models: for the early stage, the likelihood ratio χ (9) = 27.97, p < .001; for the 2 later stage, the likelihood ratio χ (9) = 42.86, p < .001. RQ1 asked whether the “traditional news organization” would be more often selected than other types of websites. The results revealed that the channel type alone was not a significant factor in either stage, indicating that in this case, “traditional news organizations” were not necessarily more popular than other channel types. H1 hypothesized that there would be a significant effect of hyperlinks on a website’s tweet frequency. The results supported H1 during both temporal stages. For the early stage, Wald χ2 (1) = 7.30, p < .01; for the later stage, Wald χ2 (1) = 5.09, p < .05. In other words, the preexisting hyperlink structure determined Twitter users’ selection of information sources. H2 hypothesized that there would be a significant interaction effect between channel types and hyperlinks. H2 was 2 supported only during the later stage, Wald χ (4) = 10.50, p < .05. A significant interaction effect was not found during the early stage. These results also supported H3, which predicted significant differences between the two time stages. AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 220 Kwon et al. Audience Gatekeeping in the Twitter Service Table 3: Tests of Model Effects of Channel Selection and Distribution in Twitter: Two-Time Period Comparison Sources Time 1 Time 2 Wald df Sig. Wald df Sig. Intercept 2.13 1 0.144 0.01 1 0.959 Hyperlinks 7.3** 1 0.007 5.09* 1 0.024 CT 2.06 4 0.725 3.79 4 0.436 CTx Hyperlinks 3.01 4 0.556 10.49* 4 0.033 Note: CT = Channel Types, * p <.05, **p < .01. Table 4 presents parameter estimates for the conditional effects of channel types and hyperlinks and their interaction effects on the later time period. When “traditional news organization” was a reference, only the parameter estimates of hyperlinks were significant. This indicates that traditional journalism websites were not the most prominent channel through which the Twitter audience consumed information during this event. When ‘social media’ was used as a reference, some parameters appeared significant, suggesting a significant influence of social media channels on the Twitter users’ gatekeeping. Accordingly, the results in Table 4 are reported using ‘social media’ as the reference level. The form of model equation for parameter estimates is as follows: log(Tweet Frequency) = Intercept + b 1 (CT=traditional) + b 2 (CT=online journalism) + b 3 (CT = personal) + b 4 (CT = org/inst/community) + b 5 Hyperlinks + b 6 (CT=traditional)*Hyperlinks + b 7 CT=(online journalism)*Hyperlinks + b 8 (CT=personal)*Hyperlinks + b 9 (CT=org/inst/community)*Hyperlinks. Table 4: Parameter Estimates on Time 2 (Social Media as a Reference Level) Parameter Hypothesis Test C.I. β S.E. Wald df Sig. Exp(B) Lower (Intercept) ** -0.96 0.74 1.69 1 0.193 0.38 0.09 1.63 (CT =Traditional) 0.68 1.01 0.45 1 0.503 1.97 0.27 14.38 Upper (CT =Online Journalism) 1.26 1.01 1.56 1 0.211 3.53 0.49 25.58 (CT =Personal Contents) 1.04 0.89 1.38 1 0.241 2.83 0.50 16.08 (CT =Org /Inst /Community) 1.74 0.94 3.41 1 0.065 5.70 0.90 36.15 Hyperlinks *** 0.25 0.07 12.84 1 0.000 1.28 1.12 1.47 (CT =Traditional) x Hyperlinks -0.12 0.10 1.39 1 0.238 0.89 0.73 1.08 (CT = Online Journalism) x Hyperlinks * -0.24 0.10 5.60 1 0.018 0.79 0.64 0.96 (CT = Personal) x Hyperlnks * -0.25 0.10 6.50 1 0.011 0.78 0.65 0.94 (CT = Org /Inst /Community) x Hyperlinks ** -0.27 0.10 6.85 1 0.009 0.76 0.62 0.93 Note: CT = Channel Types, * p < .05, ** p < .01, *** p < .001. As shown in Table 4, the interaction effect results suggest that when hyperlinks moderated the relationship between channel types and tweet frequency, “online journalism,” “personal reports,” and “organizational/institutional/community websites” were tweeted significantly less frequently than “social media.” The increase in tweet frequency of “online journalism” was 0.79 times compared to social media, β= -.24, Exp(β) = .79, p < .05; 0.78 times for “personal reports,” β = -.25, Exp(β) = .78, p < .05; and 0.79 times for “organizational/ institutional/ community websites,” β = -.27, Exp(β) = .76, p <.01. On the other hand, there was no significant difference between “traditional news organization” and “social media.” Figure 2 visually represents how the two independent variables interact with each other to determine channel popularity. To gain more information, we compared the selection pattern observed from personal users’ practices (the left column) to the comprehensive data that includes data from not only personal users but also media and other organizations and computer-automated bots (the right column). When all tweets were taken into consideration, “traditional news organization” and “social media” were the most prominently distributed types right from the beginning. Moreover, the discrepancy between these types and the rest of the content increased over time. However, when only personal users’ tweets were considered, Twitter users’ selections were not very concentrated to AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 221 Kwon et al. Audience Gatekeeping in the Twitter Service a particular media type during the early stage of news breakout. Instead, their choices were largely determined by the extent to which a website received in-links from the general online public. This observed pattern changed as the news lifespan increased, however. In the later stage, personal users’ selections were concentrated to social media-based contents, followed by the traditional news outlets. All Tweets Considered Time 2 Time1 Personal Tweets Only Note: blue = Traditional, green = Social Media, basie = Online Journalism, purple = Personal, yellow = Org/Insti/Community Sites. Figure 2: Interactions of Hyperlink Intensity and Channel Types: Comparison between ‘Personal Tweets Only’ and ‘All Tweets Considered’ Testing a Power-law Structure in Twitter Users’ Selections of News Contents RQ 2 asks if Twitter users’ selections collectively produce an unequal representation among the accessible information channels. To statistically examine the distribution structure, we tested the model of a “power-law AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 222 Kwon et al. Audience Gatekeeping in the Twitter Service distribution” following Barabasi et al.’s (2000) formalization. That is, P (k) ~ k –r, in which the probability that a website is tweeted k times follows a power law with the exponent r (Barabasi and Alberto, 1999). If there is the power-law distribution, the distribution must be represented as a negative linear relationship when plotted using log-transformed scales, which makes statistical testing possible (Moody and White, 2003); P (k) = ak –r → lnP(k) = ln(a) – r ln(k). A negative linear relationship is tested between the log-transformed value of tweet frequencies, k, and the logtransformed value of the number of websites at each k, which is P(k). Our results revealed a significant negative relationship with R2 = .80, F(1, 21) = 82.25, p < .001. In other words, the collective outcome of Twitter users’ channel selection shows an uneven distribution, suggesting the existence of a list of dominant websites. Figure 3 visualizes the imbalanced distribution of information channels. (a) A power-law distribution of the number of content websites at each tweet frequency (b) A negative linear relationship represented on a log-log plot Figure 3: Distribution of Twitter Users’ Selection of News Contents AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 223 Kwon et al. Audience Gatekeeping in the Twitter Service DISCUSSION AND CONCLUSION Scholars have attempted to theorize the role of audiences in the gatekeeping process (Barzilai-Nahon, 2008 Shoemaker and Vos, 2009) and have found that audiences play an important role in setting public agendas that is distinctly from the role played by professional journalists (Shoemaker et al., 2011). As an early attempt to apply the audience gatekeeping model to the social media context, this study investigated how Twitter audiences select information channels. An abundance of online channels can construct a news event in a variety of manners depending on the channel characteristics. Therefore, the selection of channels through which to receive and spread information can influence how the audience interprets the issue. Given that Twitter has become a popular information system that facilitates user participation for information sharing and collaboration (Honeycutt and Herring, 2009), knowing what kinds of channels are prominently selected by Twitter audiences to understand and respond to news events provides important insights. As gatekeepers, Twitter users not only consume incoming information from within their own Twitter network and/or outside of the system, but also broadcast the consumed information to their own networks of followers. We questioned how Twitter audience gatekeeping was affected by two media-centric variables: hyperlink structure and channel type. Our results revealed that while hyperlink structure largely determined users’ selections, channel type was also influential in that an interaction effect between hyperlink and channel type was present. This effect was more prominent during the later stage of news lifespan. Interestingly, the most prominent channels were social media sites, such as youtube.com, blostpot.com, and wordpress.com, rather than traditional media actors. These sites offer spaces in which the high budget products made by traditional media entities and user-generated contents intermingle. Our results suggest that the wide adoption of social media sites is a distinctive characteristic of audience gatekeeping in comparison to the whole Twitter-sphere, where traditional media actors aggressively adopt the tool as broadcasting venue. Thus, despite the prevalence of traditional mass media organizations in Twitter, audiences do not necessarily depend on them but may rather tend to find the emergent, user-empowering channels more interesting. This finding should reignite discussions about the status of mass media situated in the Web 2.0 technology era. On the one hand, proponents of citizen journalism have advocated that the participatory media would eventually challenge the “priesthood” or “punditocracy” of corporate media organizations in public information production (Chaffee and Metzger, 2001). On the other hand, a competing view has defended the professionalism of traditional news organizations regarding their investigative and global nature in contrast to the snippets and hyperlocalized stories in user-generated content (Lemann, 2006). In this study, we conclude that Twitter users’ gatekeeping process may diminish the authority of traditional organizations at least as a primary information delivery system. However, that does not mean that the audience gatekeeping in Twitter is free from existing systems. On the contrary, our results suggest that audience gatekeeping is largely determined by search results attributed to the preexisting link structure. Also, highly selected social media sites turned out to be the institutionalized commercial services. In fact, the prominent selection of a few dominant social media sites in the later stage may even be an indicator that the new media channels might continuously recycle and reconstruct the content made by big media organizations. While the popular perception is that social reporting through new technologies such as Twitter may lead to media fragmentation, our results suggest that audience gatekeeping may in fact lead to media concentration, resonant with the position of Webster and Ksiazek (2012). Meanwhile, the extent to which alternative channels are selected through audience gatekeeping should be interpreted in a mixed way. On one hand, the descriptive analysis showed that a non-trivial portion of selected channels was from alternative journalism, non-institutionalized websites, governmental, research, or non-governmental organizational databases. This result suggests that Twitter audiences utilize information alternatives to traditional news items, implying that traditional news actors may not be the predominant agenda-setters at least in the Twitter community. Also, the results showed that social media channels were as popular as traditional news channels, indicating that Twitter users adapt the items reconstructed by other social media users as much as the original news products from professional journalists. On a collective level, however, the selection pattern indicated high imbalance. The distribution revealed that Twitter was structured with a few prominent content providers – particularly the established dominant social media sites – and a large number of less visible alternative actors. The result reflects that outwardly contradictory trends coexist: one is the content concentration predisposed to a few highly prominent providers, and the other is the diversification of the news market by harnessing the long-tail part composed of less powerful actors (Anderson, 2006). The powerlaw distribution found in the current study reaffirms the recent decade’s theoretical contention between the proponents of a rich-get-richer mechanism of the Internet (e.g., Barabasi, et al., 2000; Hindman, 2008) and of the collective empowerment of newly emerging yet less visible content providers (e.g., Anderson, 2006; Barsky and Purdon, 2006; Fortunato et al., 2006). AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 224 Kwon et al. Audience Gatekeeping in the Twitter Service One limitation of our study is the generalizability of our data. The short-term collection of data with content pertaining to a long-lasting political conflict may cause history-dependent bias on user gatekeeping. In addition, the strongly political context of our data sample does not guarantee the generalizability of our findings in other social contexts, such as business. We also note that our results do not suggest any causal implications. The data could not tell us which factors preceded others in influencing user gatekeeping. For instance, preexisting reputation of traditional media may have been a significant determinant of being hyperlinked extensively. Our study exclusively focused on the pattern of channel selection, and did not delve into the reconstruction process. As gatekeepers, however, users not only distribute information but also reconstruct it with their own interpretations. Accordingly, future research needs to explore how users edit and reinterpret the existing information through their own commentary, and how the reconstruction impacts news gatekeeping in general. To construct a better understanding of audience gatekeeping, another factor that requires further investigation is preexisting channel loyalty depending on ideological predisposition, demographic, and personal attributes. ACKNOWLEDGEMENTS The authors would like to thank the SE and reviewers for their detailed feedback regarding this paper. This research has been supported by the National Science Foundation under grants 1227353 and 0916612. The research of the last (correspondent) author is also supported by the World Class University program funded by the Ministry of Education, Science, and Technology through the National Research Foundation of Korea (R31-20002) and by the Sogang University Research Grant of 2011. REFERENCES Anderson, C. (2006) The Long Tail: Why the Future of Business is Selling Less of More. Hyperion: New York. Barabasi, A., L. Albert, R. Jeong, and G. Bianconi (2000) “Power-law distribution of the World Wide Web,” Science 287, p. 2115b. Barsky, E. and M. Purdon (2006) “Introducing web 2.0: Social networking and social bookmarking for health librarians,” JCHLA/JABSC 27, pp. 65-67. Barzilai-Nahon, K. (2008) “Toward a theory of network gatekeeping: A framework for exploring information control,” The Journal of the American Society of Information Science and Technology 59 (9), 1493-1512. Berk, R. and J. MacDonald (2008) “Overdispersion and poisson regression,” Journal of Quantitative Criminology 24, pp. 269-284. Bruns, A. (2005) Gatewatching: Collaborative Online News Production. Peter Lang, New York Bruns, A. (2007) “Produsage: Towards a Broader Framework for User-led Content Creation,” in Proceedings of the 6th Creativity & Cognition Conference, Washington, DC, June 13-15, 2007. Brynjolfsson, E. and A. McAfee (2007) “The Future of the web: Beyond enterprise 2.0,” MIT Sloan Management Review 48 (3), pp. 49-55. Chaffee, S. and M. Metzger (2001) “The end of mass communication?” Mass Communication and Society 4 (4), pp. 365-379. Chang, W.-Y. (2005) “Online civic participation, and political empowerment: Online media and public opinion formation in Korea,” Media, Culture, & Society 27, pp. 925-936. Chyi, H. and M. McCombs (2004) “Media salience and the process of framing: Coverage of Columbine school shootings,” Journalism and Mass Communication Quarterly 81 (1), pp. 22-35. Clayman, S. and A. Reisner (1998) “Gatekeeping in action: Editorial conferences and assessments of newsworthiness,” American Sociological Review 68 (2), pp. 178-199. Dimitrova, D., C. Connolly-Ahern, A.Williams, L. Kaid, and, A. Reid (2003) “Hyperlinking as gatekeeping: Online newspaper coverage on an execution of the American terrorist,” Journalism Studies 4 (3), pp. 401-414. Easley, D. and J. Kleinberg (2010) Networks, Crowds, and Markets: Reasoning about a Highly Connected World. New York: Cambridge University Press. Finkelstein, S. (2008) “Google, Links, and Popularity versus Authority,” in J. Turrow and L. Tsui (Eds.) The Hyperlinked Society. Ann Arbor: The University of Michigan Press, pp. 104-124. Fortunato, S., A. Flammini, and F. Menczer (2006) “Topical Interests and the Mitigation of Search Engine Bias,” Proceedings of the National Academy of Science 103, pp. 12684-12689. Goode, L. (2009) “Social news, citizen journalism, and democracy,” New Media and Society 11 (8), pp. 1287-1305. Halavais, A (2008). “The Hyperlink as Organizing Principle,” in J. Turrow and L. Tsui (Eds.) The Hyperlinked Society. Ann Arbor: The University of Michigan Press, pp. 56-69. AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 225 Audience Gatekeeping in the Twitter Service Kwon et al. Halloran, J., P. Elliott, G. Murdock (1970) Demonstrations and Communication: A Case Study. Baltimore: Penguin Press. Hindman, M. (2008) The Myth of Digital Democracy. Princeton: Princeton University Press. Honeycutt, C. and S. Herring (2009) “Beyond Microblogging: Conversation and Collaboration via Twitter,” in Proceedings of the Forty-second Hawai’i International Conference on System Sciences, Los Alamitas, CA: IEEE Press, Hawaii, January 5-8, 2009. Israel, S. (2009) Twitterville, How Businesses can Thrive in the New Global Neighborhoods. New York: Penguin Press. Iyengar, S. (1994) Is Anyone Responsible? How Television Frames Political Issues. Chicago: The University of Chicago Press. Kang, N. and J. Choi (1999) “Structural implications of crossposting network of international news in cyberspace,” Communication Research 26 (4), pp. 454-481. Kwon, K. H. and S. Moon (2009) “The bad guy is one of us: Framing comparison between the US and Korean newspapers and blogs about the Virginia Tech shooting,” Asian Journal of Communication 19 (3), pp. 270288. Lemann, N. (August, 2006) Amateur Hour: Journalism without Journalists. New Yorker. Retrieved June 30th, 2011, from http://www.newyorker.com/archive/2006/08/07/060807fa_fact1?currentPage=all Li, J. and H. R. Rao (2010) “Twitter as a rapid response news service: An exploration in the context of the 2008 China earthquake,” Electronic Journal of Information Systems in Developing Countries, 42, pp. 1-22. Marwick, A and D. Boyd (2011) “I tweet honestly, I tweet passionately: Twitter users, context collapse, and the imagined audience,” New Media & Society 13 (1), pp. 114-133. McQuail, D. (1994) Mass Communication Theory: An Introduction, 3rd edition, London: Sage. Meraz, S. (2009) “Is there an elite hold? Traditional media to social media agenda setting influence in blog networks,” Journal of Computer-Mediated Communication 14, pp. 682-707. Mills, A., R. Chen, J. Lee, and H.R. Rao (2009) “Web 2.0 emergency applications: How useful can Twitter be,” Journal of Information Privacy and Security 5 (3), pp. 3-26. Moody, J. and D. White (2003) “Structural Cohesion and Embeddedness: A Hierarchical Concept of Social Groups,” American Sociological Review 68 (1), pp. 103-127. Napoli, P. (2008) “Hyperlinking and the Force of ‘Massification’,” in J. Turrow and L. Tsui (Eds.), The Hyperlinked Society. Ann Arbor: The University of Michigan Press, pp. 56-69. Oh, O., M. Agrawal, and H. R. Rao (2011) “Information control and terrorism: Tracking the Mumbai terrorist attack through twitter,” Information System Frontiers 13 (1), pp. 323-43. Oh, O., M. Agrawal, and H. R. Rao (2010) “Anxiety and Rumor: Exploratory Analysis of Twitter Posts during the Mumbai Terrorist Attack,” in G. Dalziel (Ed.) The Political and Social Impact of Rumor, S. Rajaratnam School of International Studies, Nanyang Technological University, Singapore. Oh, O., K. H. Kwon and H. R. Rao (2010) “An Exploration of Social Media in Extreme Events: Rumor Theory and Twitter during the Haiti Earthquake,” in Proceedings of International Conference on Information Systems, Saint Louis, MI, December 12-15, 2010. O'Reilly, T. (2005) “What is Web 2.0: Design Patterns and Business Models for the Next Generation,” Retrieved Nov 12, 2008, from http://www.oreillynet.com/pub/a/oreilly/tim/news/2005/09/30/what-is-web-20.html?page=1 Purcell, K., L. Rainie, A. Mitchell, T. Rosenstiel, and K. Olmstead (March, 2010) “Understanding the Participatory News Consumer,” Pew Internet & American Life Project. Retrieved June 30th, 2011, from http://www.pewinternet.org/Reports/2010/Online-News/Summary-of-Findings/Overview.aspx Prentice, S., P. Taylor, P. Rayson, A. Hoskins, and B. O’Loughlin (2011) “Mining author’s intent: A concept and influence analysis of terrorist media arising from the Gaza conflict,” Information System Frontiers, 13 (1), pp. 61-73. Rogers, E. (2003) Diffusion of Innovations, 5th edition. New York: The Free Press. Shin, Y.-R., and S. –H. Yim (2011, July 23) “CNN Exec Says Consumers Want News ‘on the Go’,” Korean Joonang Daily. Retrieved August 15, 2011, from http://koreajoongangdaily.joinsmsn.com/news/article/article.aspx?aid=2939260 Shoemaker, P., P. Johnson, H. Seon, and X. Wang (2011) “Readers as gatekeepers of online news: Brazil, China, and the United States,” Brazilian Journalism Research 6 (1), pp. 55-77. Shoemaker , P. and S. Reese (1996) Mediating the Message: Theories of Influences on Mass Media Content, 2nd edition. White Plains: Longman. Shoemaker, P. and T. Vos (2009) Gatekeeping Theory. New York: Routledge. Smith, A. and J. Boyle (July, 2012) “The Rise of the ‘Connected Viewers’,” Pew Internet & American Life Project. Retrieved July 31, 2012, from http://pewinternet.org/~/media//Files/Reports/2012/PIP_Connected_Viewers.pdf. AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 226 Audience Gatekeeping in the Twitter Service Kwon et al. Southwell, B. and M. Yzer (2007) “The Roles of Interpersonal Communication in Mass Media Campaigns,” in C. Beck (Ed.) Communication Yearbook 31. Mahwah, NJ: Lawrence Erlbaum, pp. 419-462. Sundar, S. and C. Nass (2001) “Conceptualizing sources in online news,” Journal of Communication 51 (1), pp. 52-72. Ward, W. (2009) “Social Media in the Gaza Conflict,” Arab Media & Society. The Middle East Center, St. Antony’s College, University of Oxford. Webster, J. and S –F. Lin (2002) “The Internet audience: Web use as mass behavior,” Journal of Broadcasting and Electronic Media, 46 (1), pp. 1-12. Webster, J., and T. Ksiazek (2012) “The dynamics of audience fragmentation: Public attention in an age of digital media,” Journal of Communication, 62(1), pp. 39-56. Yim, J. (2003) “Audience concentration in the media: Cross-media comparisons and the introduction of uncertainty measure,” Communication Monograph, 70 (2), pp. 114-128. Zanotti, J., C. Migdalovitz, J. Sharp, C. Addis, C. Blanchard, and R. Margesson (2009) “Israel and Hamas: Conflict in Gaza (2008-2009),” Congress Research Service, Retreived February 12, 2011 from http://www.fas.org/sgp/crs/mideast/R40101.pdf. Zittrain, J. (2006) “A history of online gatekeeping,” Havard Journal of Law and Technology, 19 (2), pp. 253-298. 1 A one-click function embedded in Twitter that allows a user to conveniently copy a preexisting tweet and share it via the user’s own Twitter profile. 2 http://cyberwanderer.wordpress.com/2009/01/03/summary-2009-israel-gaza/ (Retrieved February 12th, 2011). 3 As Twitter allows only 140 characters in a message to post, lengthy URLs cannot be properly posed as a linked message. To avoid this problem, many websites are providing the service of shortening the lengthy URL into a “tiny URL” format. One exemplary website is: http://tinyurl.com/. 4 The representative search engines, Yahoo.com and Google.com were removed from the analysis. AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 227 Kwon et al. Audience Gatekeeping in the Twitter Service APPENDIX The 20 Most Tweeted Web Sites in Different Time Stages Time1 Channel Name bbc.co.uk reuters.com cnn.com blogspot.com alertnet.org friendfeed.com twitter.com nytimes.com youtube.com guardian.co.uk dw-world.de aljazeera.net gazatalk.com wordpress.com haaretz.com globalvoicesonline.org israelnationalnews.com telegraph.co.uk huffingtonpost.com gmanews.tv Channel Name bbc.co.uk friendfeed.com blogspot.com alertnet.org cnn.com dw-world.de aljazeera.net guardian.co.uk gazatalk.com digg.com gmanews.tv globalvoicesonline.org allvoices.com facebook.com abc.net.au dailyme.com freealzaidi.com foxnews.com feedburner.com enduringamerica.com Tweet FRQ 67 57 36 32 32 31 29 24 21 19 19 19 13 11 11 11 11 10 10 10 Time 2 Tweet FRQ 118 71 59 52 49 41 35 33 19 17 16 16 15 14 14 13 11 11 10 9 AIS Transactions on Human-Computer Interaction Channel Types Traditional Traditional Traditional Social Media Traditional Social Media Social Media Traditional Traditional Traditional Traditional Traditional Advocacy Social Media Traditional Advocacy Traditional Traditional Online Journalism Online Journalism Channel Types Traditional Social Media Social Media Traditional Traditional Traditional Traditional Traditional Advocacy Social Media Online journalism Advocacy Online journalism Social Media Traditional Online journalism Advocacy Traditional Social Media Advocacy Hyperlinks 11.52 10.94 11.36 12.84 8.83 10.55 13.59 11.84 13.44 11.26 9.74 8.64 3.99 11.94 9.18 8.66 8.07 10.95 10.62 7.85 Hyperlinks 11.52 10.55 12.84 8.83 11.36 9.74 8.64 11.26 3.99 12.73 7.85 8.66 0.69 13.74 10.19 6.19 3.09 10.47 11.2 6 Vol. 4, Issue 4, pp. 212-229, December 2012 228 Kwon et al. Audience Gatekeeping in the Twitter Service ABOUT THE AUTHORS K. “Hazel” Kwon is an Assistant Professor in the School of Social and Behavioral Sciences at the Arizona State University. Her research interests focus on social and mobile technologies, with a particular emphasis on the dynamics in which technology-mediated communication influences collective behaviors, social networking, and civic/political participation. Her publications have appeared in Journal of Computer-Mediated Communication, CyberPsychology, Behavior & Social Networking, Computers in Human Behaviors, Asian Journal of Communication, and Journal of Information Technology & Politics. She has received Herbert S. Dordick Dissertation Award from the International Communication Association, Kappa Tao Alpha Research Award from National Honor Society in Journalism and Mass Communication, and Top Four Paper Award from National Communication Association. She completed her Ph.D. at SUNY Buffalo. Onook Oh is a doctoral candidate in the School of Management at the State University of New York at Buffalo. He is also a visiting research associate in the Center for Collaboration Science at the University of Nebraska at Omaha. His research interests are in the areas of new modalities of information exchange and social media, crowdsourcing, and use of social media in information assurance and extreme events. His papers have been published at Communications of AIS, Information Systems Frontiers, and Information Systems Management etc. He has also presented his papers at ICIS, HICCS, and other international and national information systems conferences. Manish Agrawal is an Associate Professor in the department of Information Systems and Decision Sciences of the College of Business Administration at the University of South Florida in Tampa, Florida. His current research interests include Information security, Software quality and the development of application-specific Agent-based systems. His articles have appeared in journals including Management Science, INFORMS Journal on Computing, and IEEE Transactions on Software Engineering. His research also received the Design Science Award from the INFORMS Information Systems Society. He completed his Ph.D. at SUNY Buffalo. H.R. Rao is a Distinguished Service Professor in the department of Management Science and Systems at SUNY Buffalo, USA and World Class University Visiting Professor in the department of Global Service Management at Sogang Univeristy, South Korea. His interests are in the areas of management information systems, decision support systems, e-business, emergency response management systems and information assurance. He has also received the Fulbright fellowship in 2004. He is a co-editor of a special issue of The Annals of Operations Research, the Communications of ACM, associate editor of Decision Support Systems, Information Systems Research and IEEE Transactions in Systems, Man and Cybernetics, co-Editor-in-Chief of Information Systems Frontiers and Guest Senior Editor at MISQ. Dr. Rao also has a courtesy appointment with Computer Science and Engineering as adjunct Professor. He completed his Ph.D. at Purdue University. Copyright © 2012 by the Association for Information Systems. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and full citation on the first page. Copyright for components of this work owned by others than the Association for Information Systems must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists requires prior specific permission and/or fee. Request permission to publish from: AIS Administrative Office, P.O. Box 2712 Atlanta, GA, 30301-2712 Attn: Reprints or via e-mail from ais@aisnet.org. AIS Transactions on Human-Computer Interaction Vol. 4, Issue 4, pp. 212-229, December 2012 229 Transactions on Human-Computer Interaction ISSN: 1944-3900 Editors-in-Chief http://thci.aisnet.org/ Dennis Galletta, U. of Pittsburgh, USA Ping Zhang, Syracuse U., USA Advisory Board Izak Benbasat U. of British Columbia, Canada John M. Carroll Penn State U., USA Phillip Ein-Dor Tel-Aviv U., Israel Paul Gray (Deceased) Claremont Graduate U., USA Jenny Preece U. of Maryland, USA GavrielSalvendy, Purdue U., USA, & Tsinghua U., China Ben Shneiderman U. of Maryland, USA Jane Webster Queen's U., Canada K.K. Wei City U. of Hong Kong, China Senior Editor Board Fred Davis U. of Arkansas, USA Traci Hess U. of Massachusetts Amherst, USA Shuk Ying (Susanna) Ho Australian National U., Australia Mohamed Khalifa U. Wollongong in Dubai., UAE Jinwoo Kim Yonsei U., Korea Anne Massey Indiana U., USA Fiona Fui-Hoon Nah U. of Nebraska-Lincoln, USA Lorne Olfman Claremont Graduate U., USA Dov Te'eni Tel-Aviv U., Israel Noam Tractinsky Ben-Gurion U. of the Negev, Israel Viswanath Venkatesh U. of Arkansas, USA Miguel Aguirre-Urreta DePaul U., USA Michel Avital Copenhagen Business School, Denmark Hock Chuan Chan National U. of Singapore, Singapore Christy M.K. Cheung Hong Kong Baptist University, China Michael Davern U. of Melbourne, Australia Carina de Villiers U. of Pretoria, South Africa Alexandra Durcikova U. of Arizona, USA Xiaowen Fang DePaul University Matt Germonprez U. of Wisconsin Eau Claire, USA Jennifer Gerow Virginia Military Institute, USA Suparna Goswami Technische U.München, Germany Khaled Hassanein McMaster U., Canada Milena Head McMaster U., Canada Netta Iivari Oulu U., Finland Zhenhui Jack Jiang National U. of Singapore, Singapore Richard Johnson SUNY at Albany, USA Weiling Ke Clarkson U., USA Sherrie Komiak Memorial U. of Newfoundland, Canada Na Li Baker College, USA Paul Benjamin Lowry City U. of Hong Kong, China Ji-Ye Mao Renmin U., China Scott McCoy College of William and Mary, USA Greg D. Moody U. of Nevada, Las Vegas, USA Robert F. Otondo Mississippi State U., USA Lingyun Qiu Peking U., China Sheizaf Rafaeli U. of Haifa, Israel Rene Riedl Johannes Kepler U. Linz, Austria Khawaja Saeed Wichita State U., USA Shu Schiller Wright State U., USA Hong Sheng Missouri U. of Science and Technology, USA Stefan Smolnik European Business School, Germany Jeff Stanton Syracuse U., USA Heshan Sun U. of Arizona, USA Jason Thatcher Clemson U., USA Horst Treiblmaier Purdue U., USA Ozgur Turetken Ryerson U., Canada Fahri Yetim U. of Siegen, Germany Cheng Zhang Fudan U., China Meiyun Zuo Renmin U., China 2001-2004: Ping Zhang 2004-2005: Fiona Fui-Hoon Nah 2005-2006: Scott McCoy 2006-2007: Traci Hess 2007-2008: Weiyin Hong 2008-2009: Eleanor Loiacono 2009-2010: Khawaja Saeed 2010-2011: Dezhi Wu 2011-2012: Dianne Cyr 2012-2013: Soussan Djamasbi Kar Yan Tam Hong Kong U. of Science & Technology, China Mun Yi Korea Advanced Ins. of Sci. & Tech, Korea Editorial Board Managing Editors Jian Tang, Syracuse U., USA SIGHCI Chairs http://sigs.aisnet.org/sighci