Connections to neighborhood storytellers and community-oriented emotional disclosure on Twitter during an emergency event

Abstract

From a communication infrastructure theory perspective, this study examined local residents’ connections to neighborhood storytellers and its association with community-oriented emotional disclosure, a specific form of neighborhood storytelling, during an emergency event. A sample of 1676 local residents and their Twitter activities were obtained from Twitter’s streaming application programming interface (API) during a 2-hour window after the event. A connection was established when the focal user retweeted, commented on, or replied to another user. The results show that scope of connections to residents, community organizations, and mainstream media positively predicted community-oriented disclosure of negative emotions. Scope of connections to residents positively predicted community-oriented disclosure of positive emotions. Scope of connections to local media did not predict community-oriented emotional disclosure.

Keywords

Communication infrastructure theory community-oriented emotional disclosure emergencies neighborhood storytelling Twitter

Communication infrastructure theory (CIT) provides a useful theoretical framework to understand how communities come together to solve problems collectively (Kim and Ball-Rokeach, 2006a, 2006b). CIT focuses on examining residents’ integration to the neighborhood storytelling network, which produces conversations about neighborhood events, opportunities, and problems. Many studies have found that residents who are more connected to the neighborhood storytelling network are more involved in civic life (e.g. Ognyanova et al., 2013).

Neighborhood storytellers (i.e. geo-ethnic media, community organizations, and local residents) are increasing their social media presence and forming digital layers of connections on specific platforms (An and Mendiola-Smith, 2018). However, few studies have examined storytellers’ digital traces, which could provide structural, content, and behavioral data that can be used for analysis in CIT research. Easy access to large volumes of social media data expands research opportunities for testing and modifying CIT claims. This study takes on this important task and examines the types, scopes, and effects of connections to neighborhood storytellers on Twitter. Specifically, the purpose of this study is to examine residents’ scopes of connections to neighborhood storytellers during an emergency event and how these connections are linked to their neighborhood storytelling activities.

This study makes several contributions to CIT research. This study (1) proposes and utilizes a new method to operationalize connections to neighborhood storytellers using user-generated content on Twitter; (2) tests CIT in an applied context and offers insights into the role of communication infrastructure in coping with emergencies; and (3) examines a specific form of neighborhood storytelling, namely community-oriented emotional disclosure, which is a prevalent theme in public discourse after an emergency event (Heverin and Zach, 2010; Qu et al., 2009). This article is organized as follows. Guided by the media system dependency (MSD) theory and CIT, we first review the conceptualization of connection and propose a novel way of measuring connection on Twitter. Next, we discuss the role of neighborhood storytellers in coping with emergencies and in reinforcing community-oriented emotional disclosure. Hypotheses and research questions were tested using a sample obtained from Twitter’s streaming application programming interface (API). Finally, we discuss theoretical implications and new directions for CIT research conducted on social media platforms using user-generated data.

Conceptualizing connections to neighborhood storytellers in the Context of Twitter use

CIT derives its central concept of connection from the concept of dependency relations in MSD theory (Kim and Ball-Rokeach, 2006a). MSD theory examines the interactions between audiences, media, and social systems. At the micro or individual level, MSD conceptualizes dependency relations as the extent to which the audiences rely on the resources of the media to achieve their everyday goals (Ball-Rokeach and DeFleur, 1976). In empirical studies, audiences’ dependency on a medium has been assessed by the extent to which the medium is important, helpful, or exclusive for the audiences’ attainment of goals. For example, Ognyanova and Ball-Rokeach (2015) used a cumulative index to measure Internet dependency. This index consisted of three items, each of which asked participants to rate perceived importance for the Internet to achieve the following goals, respectively—information seeking, entertainment, and maintaining social relationships. In a more recent study, Kim and Jung (2017) operationalized social networking service (SNS) dependency as the extent to which SNS is helpful in fulfilling a variety of goals (e.g. social understanding and action orientation).

CIT expands dependency relations to a wider scope of communication resources and focuses on examining the communication infrastructure of communities (Kim and Ball-Rokeach, 2006a). Connection captures local residents’ most important relations with multilevel neighborhood storytellers, who construct discourses about local residents and their community. Residents who are more connected to the neighborhood storytelling network are more likely to participate in civic activities, feel more attached to their local community, and have more confidence in community members’ ability to solve local problems together (Ball-Rokeach et al., 2001; Kim and Ball-Rokeach, 2006a, 2006b).

Residents’ needs for locally relevant information dramatically increase during an emergency event, which makes the communication infrastructure more visible (Ball-Rokeach and Jung, 2009). An emergency event is a sudden and unexpected incident that requires immediate reaction. Emergency events include, but are not limited to, earthquakes, flooding, hurricanes, volcanic eruptions, massive shootings, wildfires, nuclear explosions, and Tsunamis (American Red Cross, 2018). Studies show that Twitter is one of the most utilized social media platforms in emergency response by the public and emergency response organizations during the emergency management phase because of its capacity for quickly disseminating time-sensitive information (Van Gorp et al., 2015), whereas Facebook was preferred for day-to-day activities during the emergency preparedness phase. Twitter users are able to propagate content rapidly by writing short messages (280 character limit) and retweeting. Hashtags allow users to quickly locate information about the emergency event from users they do not follow. Residents may experience intense emotions of anxiety and fear during or after an emergency event (Tateno and Yokoyama, 2013). Intense emotions are usually short-lived (Scherer, 2005). Twitter is one of the few widely used social media platforms that publicly document users’ immediate emotional responses to emergencies (Heverin and Zach, 2010). A number of studies have examined how the public reacted to various emergency events (Acar and Muraki, 2011; Qu et al., 2009; Vieweg et al., 2010) and how government agencies and organizations used Twitter for emergency management (Genes et al., 2014; Van Gorp et al., 2015).

In this study, we consider that a dependency relation is established when a user takes an explicit action on a tweet, which includes retweeting, quoting, or replying to that tweet. This set of interactive communication is indicative of an important connection between the focal person and the source for several reasons. First, when users are engaged in these actions, the content and/or author of another tweet appears on the focal user’s timeline, suggesting the involvement in public conversations on social media. The inclusion of such tweets on someone’s timeline is a deliberate process; that is, users take cognitive effort in evaluating the content and source of the tweets and selectively re-post tweets that are meaningful to them (Counts and Fisher, 2011). Starbird and Palen (2010) found that Twitter users retweeted information they perceived as important for others to know in the event of an emergency. Twitter users also retweet when they want to publicly validate and agree with someone’s thoughts (boyd et al., 2010).

Second, writing demands more cognitive effort than reading (Piolat et al., 2005). In an experiment, Counts and Fisher (2011) asked participants to read their Twitter timelines and used eye-tracking techniques to measure time spent on reading each tweet. On average, participants spent 2.92 seconds on reading one tweet. Tweets would be more likely to be replied to or retweeted when users spent more time reading them, considered them interesting, and remembered them better. The results suggest that when users take an explicit action on a tweet, they need to be attentive to the tweet and allocate resources of their working memory to carefully process and respond to the tweet. It is unlikely that users spend their precious time responding to tweets that are meaningless or unimportant to them.

Finally, mentions and retweets have often been used as metrics of user influence. Cha et al. (2010) viewed retweet and mention as indicative of “the ability of that user to generate content with pass-along value” and “the ability of that user to engage others in a conversation,” respectively (p. 12). From the perspective of the user who retweets or mentions another tweet, it can be inferred that this user receives influence from another user, indicative of an important relationship between the two users.

Neighborhood storytellers during an emergency event

The neighborhood storytelling network is composed of three key storytellers, including geo-ethnic media, community organizations, and local residents (Kim and Ball-Rokeach, 2006b). Geo-ethnic media target specific geographical areas and/or ethnic populations. They play an important role in reaching ethnic minorities who reside in culturally diverse communities (Matsaganis et al., 2010). Starbird and Palen (2010) examined retweeting behavior after two emergency events, including the Red River flooding in North Dakota and Minnesota and the Oklahoma wildfires, in Spring 2009. It was found that some of the most retweeted tweets originated from local media who provided specific and relevant information about the events.

Community organizations take a variety of forms ranging from informal neighborhood associations to formal non-profit organizations (Kim and Ball-Rokeach, 2006b). Research shows that emergency response organizations have used social media for emergency management in a number of ways. Van Gorp et al. (2015) examined social media use among 50 emergency response organizations. These organizations used social media to disseminate locally relevant information and to build collaborative work with their partners. A report published by the US Department of Homeland Security (2013) reveals that emergency response organizations used social media to gather information from the public who posted photos and videos of the affected areas. This information was used by the response team to plan their response efforts. During Hurricane Isaac, the New Orleans mayor’s Twitter account was used directly to respond to community concerns and to correct misinformation (US Department of Homeland Security, 2013).

Local residents and their family members, friends, and neighbors are essential micro-level neighborhood storytellers. During emergencies, local residents post firsthand knowledge of the event, update safety status, express opinions and emotions, and correct misinformation (Acar and Muraki, 2011; Heverin and Zach, 2010; Mendoza and Poblete, 2010; Qu et al., 2009).

Recent CIT research explores a larger communication ecology, which goes beyond the neighborhood storytelling network. A communication ecology approach focuses on how residents use a web of communication resources to achieve specific goals (Ball-Rokeach et al., 2012). In addition to the three key storytellers described in CIT, communication ecology also includes storytellers, such as social media, mainstream media, non-profit organizations that have a national or international clientele, and formal interpersonal connections (Wilkin, 2013). For example, residents residing near the Fukushima Daiichi Nuclear Plant explosion site actively searched for information about risks of radiation exposure and radiation protection from academic scholars, medical doctors, and mainstream TV channels and newspapers (Tateno and Yokoyama, 2013).

Studies (Kim and Jung, 2017; Ognyanova and Jung, 2017) that examine the role of emerging technology in neighborhood storytelling networks have found that technology could both be integrated into the existing offline networks to complete similar tasks (technology as a facilitator) and be drivers of change for new tasks (technology as a catalyst). In this study, we take the perspective of technology as a facilitator and view that Twitter provides a new platform to facilitate community engagement. Research shows that people are likely to use different channels that perform similar functions to achieve the same goal (Dutta-Bergman, 2006). For example, Smith (2013) found that the majority of users who are actively engaged in political and civic activities on social media platforms are also involved in a wide range of behaviors outside of these platforms, such as attending a political meeting, expressing their opinion via an email, or signing a paper petition. We also view that storytellers’ Twitter use forms one digital layer of connections that constitute a subset of the broader connections in the neighborhood storytelling network. It is partly because only a subset of neighborhood storytellers have an active presence on Twitter. These storytellers tend to be residents who have higher levels of education and income (Smith, 2013) and organizations that have a broader geographical clientele and the capacity to manage their social media accounts (An and Mendiola-Smith, 2018). With this in mind, we will examine what types of storytellers local residents are mostly connected to on Twitter during an emergency event:

RQ1. Who are the major storytellers on Twitter during an emergency event?

Community-oriented emotional disclosure as neighborhood storytelling

According to CIT, neighborhood storytelling refers to “any type of communicative action that addresses residents, their local communities, and their lives in those communities” (Kim and Ball-Rokeach, 2006a: 178). Broadly defined, neighborhood storytelling involves discussions on topics such as local events and activities, housing development, after school programs, crime and safety, and traffic congestion. Neighborhood storytelling takes various forms but has to be about the local community (Kim and Ball-Rokeach, 2006a). Emotional disclosure refers to the process of communicating the emotion-eliciting event and emotional responses (Rimé, 2009). We view emotional disclosure as a form of neighborhood storytelling when residents explicitly write about the emergency event and their emotional responses elicited from the event. We use the term community-oriented emotional disclosure to highlight the local context bounds of this special type of emotional disclosure. When residents share their emotions on Twitter, they transform private feelings to public discourses and engage in public conversations about the event. Residents’ disclosed emotions may be acknowledged and responded to by their existing ties on social media as well as by other residents who use hashtags or keywords to locate posts containing community-oriented emotional disclosure. Community-oriented emotional disclosure thus constitutes neighborhood storytelling, which contributes to the collective construction of emotional experiences (Rimé, 2009). In addition, neighborhood storytelling is the act of constructing an identity as a member of the neighborhood (Ball-Rokeach et al., 2001). An emergency event that has clear geographical boundaries may activate local residents’ collective identity since they may wonder how other members of the neighborhood are affected by and respond to the event. Thus, community-oriented emotional disclosure about the shared event is not an isolated act, but a process of identity construction as a member of the neighborhood. Disclosure of emotions that are irrelevant to the emergency event is not considered neighborhood storytelling in this study since such disclosure may not address community issues.

The community-oriented disclosure of emotions elicited from emergency events is prevalent on social media. For instance, Qu et al. (2009) examined comments from an online community in response to the 2008 Sichuan earthquake. One common theme was that members of this community expressed negative emotions such as anxiety, sadness, and anger. Another study collected public tweets relating to the shooting of four police officers in Seattle, which found that some users expressed feelings of fear (Heverin and Zach, 2010). Positive emotions, such as gratefulness and blessing, have also been found in online communities after emergency events (Heverin and Zach, 2010; Qu et al., 2009).

Research on emotional disclosure suggests that disclosure of negative emotions and disclosure of positive emotions may have different mechanisms in which individuals cope with traumatic events. Traumatic events pose threats to the self-concept, which tend to make people repetitively think about details, causes, and consequences of these events (Rimé et al., 1998). When individuals pay excessive attention to themselves, they may disassociate themselves from the social environment. Disclosure of negative emotions helps individuals restore beliefs about themselves, clarify ambiguous sensations, reconnect them with the social environment, and interpret the event in a meaningful way. Disclosure of negative emotions also helps individuals solicit support, advice, and empathy from their social networks (Rimé et al., 1998). For example, Wang et al. (2015) content analyzed 1000 thread starting messages and their first replies in an online breast cancer support community and found that disclosure of negative emotions was positively associated with receiving emotional support from the community. Disclosure of positive emotions after an upsetting event is as therapeutic as disclosure of negative emotions (Segal et al., 2009). Negative emotional events fuel cognitive work (Rimé, 2009). Disclosure of positive emotions after a traumatic event may be indicative of cognitive reappraisal, through which individuals transform negative emotional responses to positive interpretations. Cognitive reappraisal is an essential coping strategy to alleviate emotional distress (Burleson and Goldsmith, 1998). Furthermore, disclosing positive emotions is a widely used strategy to manage impressions on social media (Gil-Lopez et al., 2018; Lin et al., 2014). Disclosure of positive emotions after traumatic events may help individuals maintain a desired self-image by being perceived as strong, hopeful, and positive.

A basic premise of CIT is that “local communities are based on resources for storytelling about the community” (Kim and Ball-Rokeach, 2006a: 177). CIT research focuses on examining the extent to which residents are connected to these resources. Recent research on online communities utilizes the structural approach for resource assessment. The structural approach examines users’ structural positions (e.g. number of friends/followers) within a social network, which are indicative of their capacity to acquire resources embedded in the network (Burt, 2004). This approach separates structure of relations from resources/outcomes produced by the structure. In the context of CIT, we view residents’ connections to neighborhood storytellers as their structural positions, which provide opportunities for residents to engage in a variety of activities (e.g. neighborhood storytelling). For example, Ball-Rokeach et al. (2001) found that the scope of connections to community organizations and scope of connections to local media positively predicted residents’ intensity of participation in neighborhood storytelling. In this study, we follow this line of argument and examine how local residents’ structural connections to neighborhood storytellers are linked to their production of neighborhood discourses. We focus on a specific form of neighborhood storytelling and argue that the scope of connections to the neighborhood storytelling network is positively linked to community-oriented emotional disclosure among residents who experience an emergency event. When a person perceives that a feeling, thought, or action is shared by influential others on social networks, he or she develops the attitudinal evaluation of that feeling, thought, or action, which is congruent with that of the others (Friedkin, 2001). During an emergency event, residents are surrounded by stories of the incident. The shared community discourses normalize feelings of the residents and project an expectant audience who is able to understand what they have been going through. Connections to these discourses may make residents feel secure and reinforce their own disclosing behavior. In addition, the dyadic effect describes reciprocity—the process by which people take turns to reveal personal information or feelings to one another (Altman, 1973). Research shows that when one member of an online community self-discloses, another member is likely to reciprocate self-disclosure (Barak and Gluck-Ofri, 2007). Community-oriented emotional disclosure about an emergency event from neighborhood storytellers may elaborate and validate residents’ emotional experience and encourage self-disclosure of their own emotions. Therefore, we hypothesize the following:

H1. Scope of connections to neighborhood storytellers will be positively associated with community-oriented disclosure of negative emotions on Twitter.

H2. Scope of connections to neighborhood storytellers will be positively associated with community-oriented disclosure of positive emotions on Twitter.

Method

Study context

On 13 January 2018, at 8:07 a.m., the Hawaii Emergency Management Agency (HEMA) sent out a text alert reading: “BALLISTIC MISSILE THREAT INBOUND TO HAWAII. SEEK IMMEDIATE SHELTER. THIS IS NOT A DRILL.” The message was received on cell phones of residents and visitors alike as the system uses geolocation to send the alerts. The warning was also broadcast over television and radio, and some residents and visitors reported hearing the state warning system sirens located around the coast, which are routinely tested and used for tsunami warnings. The HEMA website crashed shortly after the message was released and was not operational again until around 9:30 a.m. Flights at Honolulu International Airport were suspended for 18 minutes. According to the Honolulu Police Chief, Susan Ballard, 911 dispatchers answered around 3000 calls after the alert was issued and they estimated that another 2500 calls were dropped (Hawaii News Now, 2018). Minutes after the alert was released, the US Pacific Command issued the following statement on Twitter: “U.S. Pacific Command has detected no ballistic missile threat to #Hawaii. Earlier message was sent in error and was a false alarm.” At 8:20 a.m., HEMA issued a retraction via its Twitter account stating, “NO missile threat to Hawaii.” Hawaii’s Governor, David Ige, also issued a similar statement on Twitter at 8:24 a.m. saying, “There is NO missile threat.” At 8:45 a.m. HST, 38 minutes after the initial alert was sent, a second emergency alert was sent to cell phones by HEMA, which stated, “There is no missile threat or danger to the State of Hawaii. Repeat. False Alarm.”

Data collection

A connection request was sent to Twitter’s Streaming API approximately an hour after the emergency alert. Five filter words were used, including “Hawaii,” “alert,” “ballistic missile,” “threat,” and “emergency.” Once the connection was established, the stream sent in real-time tweets containing the filter words. Data collection lasted 2 hours and the returning JSON object contained 262,547 tweets. Non-English and duplicated tweets were removed, which resulted in a subset of 243,273 tweets. Since we were interested in users’ original and deliberate construction of neighborhood discourse, we removed users who only retweeted without a comment. The remaining sample contained 92,712 tweets by 46,314 unique users.

A list of 155 location keywords was compiled. The list contained names of the counties, cities, towns, and villages of Hawaii (List of places in Hawaii, 2018). “Hawaii” and “HI” were also added to the list. A user was classified as local if his or her profile location contained at least one keyword from the list. A total of 1752 users were local users, 33,082 were non-local users who specified a location but did not use any keyword from the list in their profiles, and 11,480 users did not specify a location in their profiles.

Measures

Scope of connections

The following explains different types of tweets. A general tweet is an original message posted on Twitter. An example is, “Hawaii just got memed so hard.” A retweet is a forwarded message without comments. The format is, “RT @username there is an emergency alert.” The username of the original tweet can be found in the “retweeted_status” object. An example is, “RT @Hawaii_EMA: NO missile threat to Hawaii.” A quote tweet is a forwarded message with the user’s own comment added. The format is, “comment https://. . ..” The content of the original tweet is not displayed, instead, the URL of the original tweet is added to the comment. The username of the quoted tweet can be found in the “quoted_status” object. An example is, “My heart couldn t take 40 minutes. https://t.co/1PYHIebDDt.” When a quoted tweet is being retweeted, the URL of the original tweet is also displayed. An example is, “RT @brianschatz: Please retweet https://t.co/ry6FPmUQNS.” In this example, the URL is the address of the original tweet (a), user “brianschatz” retweeted tweet (a) and added his comment “Please retweet,” which formed tweet (b). Tweet (b) was retweeted by a third user without adding comments and formed tweet (c), which is the above example. A reply is when the user responds to another person’s tweet. The format is @username comment. A reply does not display the content of the original tweet. The “in_reply_to_screen_name” variable of the JSON file points to the username being replied to. An example is, “@maziehirono what a terrible thing for the people of Hawaii can’t imagine the terror.” A mention, which contains another user’s Twitter username, has a similar format. The username of a mention is preceded by the “@” symbol. The “entities.user_mentions” object lists all the users being mentioned in the tweet. Although replies and mentions appear to have similar format, the nature of the connection can be differentiated by extracting appropriate objects in the JSON file.

The type of storytellers was coded based on users’ profiles (1 = local media, 2 = community organizations, 3 = resident, 4 = mainstream media, 5 = alternative media, 6 = celebrities, and 7 = business). Two coders independently classified 1106 users whose tweets had been commented on, retweeted, and replied to by local residents. Local media (n = 24) were operationalized as media outlets serving residents who live in Hawaii and individuals who claim to be affiliated with local media (e.g. editor, journalist, host). Examples include Hawaii News Now, KHON News, Star-Advertiser, and KITV4. Community organizations (n = 32) were operationalized as politicians, government agencies, and nonprofit organizations. Examples include Hawaii Emergency Management Agency, mayor of the City and County of Honolulu, Hawaii’s Governor, Hawaii Red Cross, Honolulu Police, and Office of Hawaiian Affairs. Mainstream media (n = 131) were operationalized as media outlets that have an international, national, and regional audience as well as individuals who claim to be affiliated with these media outlets. Examples include CNN, Fox News, Time, ABC, BuzzFeed, CBS News, Business Insider, and Fortune magazine. Alternative media (n = 106) include accounts that have more than 100,000 followers but were not classified as mainstream media. Examples include YouTube personalities, independent bloggers, news websites, and activists. Celebrities (n = 37) include actors, producers, and other public figures (e.g. CEO of a multinational corporation). Businesses (n = 10) were operationalized as users whose profile contained a business mission statement or address, such as shops, restaurants, and wireless providers. Users who were not classified as the above categories were classified as residents (n = 766)^.1

The scope of connections is operationalized as the number of unique users for each type of storyteller whose tweets are being retweeted, commented on, or replied to by the focal user. For example, Resident A’s scope of connections to local media would be 2 if A retweeted local media B’s tweet and commented on local media C’s tweet. Resident A’s scope of connections to other local residents would be 1 if A retweeted Resident B’s first tweet and also commented on B’s second tweet. This study’s operations of connections to neighborhood storytellers were different from those that were measured in previous CIT survey studies, which asked residents to report types of news sources used to stay informed about their community, organizational membership, and frequency of interpersonal discussion (e.g. Ognyanova et al., 2013). This study’s operations resulted in a selective and context-specific inclusion of neighborhood storytellers. For example, this study captured community organizations that were mainly political and emergency-related and did not include those that pursued cultural, educational, or other public-benefit goals.

Community-oriented emotional disclosure

A total of 2616 tweets were posted by local residents. These tweets had at least some forms of the user’s original expression, such as comments on another tweet/retweet, replies to other users, or general tweets. Two coders independently reviewed these tweets and coded whether each tweet contained community-oriented disclosure of negative emotions (1 = present, 0 = absent) and community-oriented disclosure of positive emotions (1 = present, 0 = absent). Cues that were used to detect emotions include emotion words (e.g. “scared,” “panic,” “glad,” and “thankful”), emotional punctuation (e.g. “?!!!”), acronyms (e.g. “lol”), and strong language (e.g. “I literally moved from Hawaii this month back to Scotland due to this.”). Tweet content was coded at face value. No contextual information was used to code the presence of emotions. Two rounds of coding and discussions were performed. Discrepancies were discussed until agreement was reached for all tweets. An index was calculated by summing up the number of tweets containing community-oriented disclosure of negative emotions for each user (range 0–9, M = .60, standard deviations [SD] = .85). A second index was calculated by summing up the number of tweets containing community-oriented disclosure of positive emotions for each user (range 0–8, M = .24, SD = .53).

Community-oriented disclosure of negative emotions includes comments that described anxiety and fear (e.g. “I live in Hawaii it was terrifying!” “We were so scared . . . People were in tears!” “I woke up scared, panicked, and wondering what was going on.” “I am still shaking . . . Scariest moment of my life.”), comments that expressed anger and dissatisfaction (e.g. “I’m really angry. This is NOT a Reality show. This is our lives!” “I am extremely disappointed with the unnecessary panic that the people in the state of Hawaii and their loved ones received this morning.”), comments that conveyed criticisms and loss of trust (e.g. “So tired of the corrupt Hawaiian government. Can’t even trust them to manage our early warning center.” “Kona here unacceptable delay in advising people that this was a false alarm by the emergency response system and media! No confidence in emergency warning system.”), and comments that showed sadness (e.g. “So sad this happened this morning here in Hawaii.” “Sadly, this is yet another example of Hawaii’s incompetence in technology.”).

Community-oriented disclosure of positive emotions includes comments that expressed relief (e.g. “One of those mornings where your life flashes before you, but such a relief. Just reminds me how much I should let my loved ones know I care.” “So glad the missile didn’t hit for real . . .”), thankfulness (e.g. “So so grateful for and blessed by all of the loving texts and calls I received from loved ones ~ both in Hawaii and around the world” “Aloha Tulsi, you care for your people of Hawaii, took immediate action and informed us first, your heart is always here!”), and hopeful/cheerful states (e.g. “Its a beautiful and safe day in Hawaii. Enjoy your weekend.” “Proud of my wife. She kept the elderly community safe and calm and put every one of those people before herself during this hectic morning.”).

A tweet could contain both community-oriented disclosure of positive emotions and community-oriented disclosure of negative emotions. An example is, “In the case that this alert was real, I am impressed with the ability to notify the state on a mass level so quickly. However, I am extremely disappointed with the unnecessary panic that the people in the state of Hawaii and their loved ones received this morning.” The first part of the tweet “I am impressed with the ability to . . .” was coded as community-oriented disclosure of positive emotions, and the second part of the tweet “However, I am extremely disappointed with . . .” was coded as community-oriented disclosure of negative emotions.

Analysis

Two Poisson regression models were constructed using R 3.5.1 to examine the relationship between scope of connections to storytellers and community-oriented emotional disclosure among local residents. The independent variables were identical for the two models, including three control variables (i.e. number of followers, number of friends, and number of statuses) and connections to each type of neighborhood storyteller (i.e. local media, community organizations, residents, mainstream media, alternative media, celebrities, and businesses). The dependent variable was community-oriented disclosure of negative emotion for the first model and community-oriented disclosure of positive emotion for the second model.

Results

The local sample consisted of 1676 local residents who posted, on average, 1.56 tweets that contained some forms of their original expression. RQ1 examined who the major storytellers were on Twitter during an emergency event. The means and SD of scope of connections to storytellers were as follows: local media (M = .10, SD = .34), community organizations (M = .20, SD = .51), residents (M = .67, SD = 1.50), mainstream media (M = .41, SD = 1.09), alternative media (M = .15, SD = .50), celebrities (M = .05, SD = .24), and businesses (M = .01, SD = .09). Table 1 presents the correlations among the variables.

Table 1.

Zero-order correlations among variables.

	1	2	3	4	5	6	7	8	9	10	11	12
1. Negative	1
2. Positive	.11^***
3. Follower count	−.02	−.01
4. Friend count	.00	−.02	.79^***
5. Status count	.04	.01	.20^***	.19^***
6. Local media	.09^**	−.01	−.01	.02	.10^**
7. Organizations	.23^***	.04	.00	.02	.08	.17^***
8. Residents	.20^***	.11^***	.00	.01	.13^***	.15^***	.23^***
9. Mainstream	.25^***	.00	.00	.02	.15^***	.22^***	.32^***	.33^***
10. Alternative	.13^***	.04	.00	.00	.08*	.11^***	.18^***	.32^***	.4^***
11. Celebrity	.13^***	.02	.00	.01	.04	.04	.13^***	.14^***	.22^***	.15^***
12. Business	.03	.00	−.01	.00	−.01	.03	−.02	.01	.04	.05	.01	1

p < .05, **p < .01, ***p < .001.

H1 predicted that scope of connections to neighborhood storytellers would be positively associated with community-oriented disclosure of negative emotions on Twitter among local residents. The model was statistically significant, with $λ_{(10)}^{2}$ = 136.6, p < .001. As shown in Table 2, connection to community organizations (b = .28, p < .001), connection to residents (b = .05, p < .001), connection to mainstream media (b = .08, p < .001), and connection to celebrities (b = .25, p < .05) positively predicted community-oriented disclosure of negative emotions. Local residents who retweeted, commented on, or replied to one more community organization were 1.32 (e^.28) times more likely to disclose negative emotions, while holding all other variables constant in the model. The estimated rate ratios were 1.05 (e^.05), 1.08 (e^.08), and 1.28 (e^.25) for residents, mainstream media, and celebrities, respectively. H1 was partly supported.

Table 2.

Regression analysis predicting community-oriented emotional disclosure.

	Emotional disclosure
	NegativeBeta (SE)	PositiveBeta (SE)
Control
Follower count	−.11 (.09)	.01 (.06)
Friend count	.08 (.08)	−.07 (.13)
Status count	−.00 (.03)	.01 (.05)
Connection to storytellers
Local media	.00 (.08)	−.02 (.17)
Community organizations	.28^*** (.05)	.15 (.09)
Residents	.05^*** (.01)	.10^*** (.02)
Mainstream media	.08^*** (.02)	−.10 (.06)
Alternative media	.00 (.06)	.04 (.10)
Celebrities	.25^* (.10)	.06 (.20)
Businesses	.45 (.29)	.05 (.58)
Pseudo-R²	.12	.02
Null deviance (df)	1825.8 (1675)	1311.1 (1675)
Residual deviance (df)	1689.2 (1665)	1288.4 (1665)

SE: standard error; df: degrees of freedom.

Follower count, friend count, and status count were standardized variables. Nagelkerke R² was reported as Pseudo-R².

p < .05, ***p < .001.

H2 predicted that scope of connections to neighborhood storytellers would be positively associated with community-oriented disclosure of positive emotions on Twitter among local residents. The model was statistically significant, with $λ_{(10)}^{2}$ = 22.7, p < .05. As shown in Table 2, connection to residents (b = .10, p < .001) positively predicted community-oriented disclosure of positive emotions. Local residents who retweeted, commented on, or replied to one more resident were 1.11 (e^.10) times more likely to disclose positive emotions, while holding all other variables constant in the model. H2 was partly supported.

Discussion

This study examined local residents’ connections to the neighborhood storytelling network in the event of an emergency, during which a high need for information makes the local communication infrastructure more visible. Since Twitter has been commonly used by neighborhood storytellers to quickly disseminate information about events (Van Gorp et al., 2015), this study focused on connection patterns on Twitter, which constituted one digital layer of all connections within the broader neighborhood storytelling network. The results indicate that scope of connections to key storytellers positively predicted community-oriented emotional disclosure—a specific form of neighborhood storytelling.

The results of this study show that local residents were connected to a wide range of storytellers who shaped the community discourse about the emergency event. In terms of scope of connections (as measured by the number of unique users whose tweets are being retweeted, commented on, or replied to by the focal user), residents were most connected to other residents, followed by mainstream media, community organizations, alternative media, local media, celebrities, and businesses. Consistent with previous studies, residents largely rely on interpersonal networks for local news and information (e.g. Ognyanova et al., 2013) regardless of study contexts. Tweets posted by mainstream media, such as Fox News, CNN, MSNBC, were being most retweeted, commented on, or replied to by different local residents. By contrast, local media had a relatively weak presence on Twitter. Because of the nature of the event, the vast majority of community organizations were politicians and government agencies with only a few exceptions, including Hawaii Red Cross, Pacific Command, the New Agenda (an advocacy group for women and LGBTQ), Hawaii Red Cross, Office of Hawaiian Affairs (oha.org), and Democrats Work For America. This study captured more than 100 alternative media users, such as YouTube personalities, independent bloggers, news websites, and activists, to which local residents were connected for emergency-related information. It suggests that CIT research may incorporate these social media- or web-based storytellers into the communication ecology.

An important finding of this study is that interpersonal networks play a central role for the public in coping with the emergency event. Scope of connections to residents was the highest among that of all neighborhood storytellers. It was the only factor that predicted community-oriented disclosure of both negative emotions and positive emotions. Research shows that information disseminated via peer-to-peer relations is more relevant and personal (Starbird and Palen, 2010; Veil et al., 2011). It is possible that residents’ tweets contained detailed accounts of their physical and emotional reactions to the event, which validated and normalized the feelings of the focal person. This validation conveyed a sense of social support, which might encourage residents’ disclosure of undesirable feelings. This finding supports CIT but contradicts other studies reporting that individuals were less likely to disclose negative emotions when they had more contacts on social media because disclosing negative emotions might undermine self-image (Gil-Lopez et al., 2018). The difference may be explained by the different approaches in conceptualizing emotional disclosure, either as revealing personal events or as narrating shared events. Furthermore, impression management might be a plausible explanation for the relationship between scope of connections to residents and community-oriented disclosure of positive emotions. Lin et al. (2014) found that the more friends Facebook users had the more likely these users disclosed positive emotions on Facebook. Disclosing positive emotions may convey a favorable self-image to others such as being grateful, hopeful, and caring. It is also possible that residents who were more connected to other residents received more social support, and thus were more likely to publicly acknowledge their received support. Finally, communities in Hawaii are characterized by close-knit ohana (family) and friends. A previous study shows that individuals were more likely to engage in self-presentation when their social media networks were large and homogeneous (Rui and Stefanone, 2013). It is possible that Hawaii residents felt accepted to self-disclose community-oriented negative emotions and tried to be supportive when self-disclosing community-oriented positive emotions.

Community organizations were residents’ third most connected neighborhood storyteller in this study. These organizations include emergency response organizations and government officials. This result is consistent with previous studies (US Department of Homeland Security, 2013), indicating the public’s need for information from opinion leaders to understand what has happened. These opinion leaders are insiders or decision makers who have access to credible information on the event. The information can be helpful for local residents to cope with ambiguity and anxiety. Furthermore, this study found that scope of connections to community organizations positively predicted community-oriented disclosure of negative emotions. The strength of this relationship was stronger than that of the relationship between connections to local residents and community-oriented disclosure of negative emotions. It may be that the public held the government accountable and vented their negative feelings toward those who caused panic. Post hoc observations showed that the tone of the tweets posted by community organizations appeared to be neutral and factual (e.g. “No missile threat to Hawaii.”), as compared to tweets from interpersonal networks that conveyed emotions and opinions (e.g. “So we almost had state panic . . .,” “Our leaders have failed us . . .”). It is possible that the neutrally toned tweets did not affirm residents’ emotions elicited from the incident, and thus, might increase residents’ likelihood of disclosing negative emotions to justify their emotional experiences. Previous CIT studies have found limited role of community organizations in foster online neighborhood storytelling and community-oriented online participation (An and Mendiola-Smith, 2018; Ognyanova et al., 2013). To the contrary, the findings of this study highlight the important role of community organizations in fostering neighborhood storytelling, suggesting the context-specific roles of community organizations in the neighborhood storytelling network. Consistent with findings from a previous study (An and Mendiola-Smith, 2018), this study identified organizations that tend to have a wider clientele and resources for managing social media activities and did not find social media presence for informal groups that have a small-scale clientele.

Scope of connections to mainstream media was significantly associated with community-oriented disclosure of negative emotions. A few studies have reported that news stories from mainstream media exaggerated insignificant incidents and used emotional appeals to obtain attention from audiences during emergency events (Heinzelman and Waters, 2010). It may be that exaggerated news stories propagated panic, which reinforced and strengthened negative emotions experienced by local residents. Furthermore, scope of connections to local media did not predict community-oriented emotional disclosure. It is possible that the influence of local media was primarily offline rather than online. The results of this study show that the majority of local news outlets had few connections with local residents except two local television stations—Hawaii News Now and KHON2 News. Other local news outlets, such as newspapers, radio stations, and television stations, had no or limited presence on Twitter during the emergency event. These traditional media outlets were more appealing to certain populations (e.g. older adults) who were the least likely to use social media to stay on top of what was happening in the community (Pew Research Center, 2017; Smith and Anderson, 2018).

Implications for CIT research

The role of technology in CIT research has been evolving. A recent trend is a shift from general patterns of technology use to platform-specific uses (Ognyanova and Jung, 2017). The availability of social media data in large volumes and scopes not only provides new opportunities for testing CIT on specific social media platforms but also poses challenges to operationalizing key CIT constructs. First, this challenge is partly caused by the missing piece of information on pursuing goals associated with user activities. This information has been clearly stated in grounded community research (e.g. surveys and focus groups). The present study tried to tackle this challenge by examining communication activities after an emergency event, when residents in the affected neighborhoods shared a common goal—to stay on top of the emergency event. CIT research could be advanced by focusing around major community events or using features of social media platforms (e.g. hashtags) to extract goal-oriented activities. Second, this study is one of the first attempts to operationalize connection by tracing user activities on Twitter. This operationalization made non-intrusive research possible for CIT studies conducted on social media or other online platforms. Future CIT research could use web logs and/or other user activity data to identify important dependency relations in the neighborhood storytelling network. The behavioral data might be more accurate and subtle than data obtained from retrospective surveys. An additional advantage of using behavioral data to operationalize connection is that researchers could examine cross-level dependency relations simultaneously. Example dependency relations include how local media depend on residents to achieve their professional goals (e.g. information gathering) and how community organizations depend on local media to achieve their organizational goals (e.g. community outreach). Future CIT research could examine how the scope, intensity, and power differentials of connections among key storytellers exert influence on neighborhood storytelling and civic engagement. Finally, the large volume of textual data document detailed accounts of neighborhood storytelling and residents’ sentiments, which may provide opportunities for understanding the nuanced paths to neighborhood belonging. For example, this study selected one aspect of neighborhood storytelling, namely community-oriented emotional disclosure. Emotional disclosure promotes closeness in interpersonal relationships (Rimé, 2009). Community-oriented emotional disclosure as neighborhood storytelling transforms residents’ private feelings to shared discourse, which may promote closeness among neighborhood members. Community-oriented disclosure of negative emotions may increase residents’ access to community resources (e.g. social support) when coping with emergencies. Community-oriented disclosure of positive emotions may reinforce the positive thinking of being associated with the community. Future CIT research could examine the content and features of neighborhood storytelling and their paths to neighborhood belonging.

Limitations

This study has some limitations. First, the operationalization of connection only captured a subset of dependency relations because a dependency relation could also exist when a user read a tweet to understand his or her social environment, but did not retweet, quote, or reply to that tweet. This type of dependency relation was excluded in this study due to data unavailability, such as time spent on reading a tweet, when collecting public real-time tweets. Future studies could use web log data or other types of data to capture a complete set of dependency relations within the neighborhood storytelling network. Second, data collection and cleaning posed some sampling challenges. This study may suffer from common sampling bias in that Twitter users tend to be younger and have higher levels of education and income (Smith and Anderson, 2018). In addition, this study used location keywords to filter out non-local residents. Those who resided in Hawaii but did not reveal their residence locations on Twitter biography page were excluded from the final sample. Third, this study only coded the textual component of residents’ tweets. Other components of tweets, such as images and webpage URLs, were not coded but might contain information about residents’ emotional disclosure. Finally, we coded residents’ global positive or negative emotional states because we found emotion blends (e.g. feeling both anxious and angry) in the tweets. Also, we did not have a sufficient sample to conduct a meaningful analysis on specific emotions (e.g. fear, anxiety, and anger). Future studies could utilize computerized text analysis tools to accurately classify specific emotions on large samples and gain a nuanced understanding of emotional disclosure during emergency events.

Footnotes

Acknowledgements

The authors would like to thank the editor and the two anonymous reviewers for their helpful feedback.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the University of Hawai‘i at Hilo.

ORCID iD

Zheng An

Notes

Author biographies

Zheng An (PhD, University of Southern California) is an assistant professor in the Department of Communication at the University of Hawai‘i at Hilo. Her research examines civic engagement, social support, and social networks.

Luana Mendiola-Smith is an undergraduate communication studies major student at the University of Hawai‘i at Hilo.

References

Acar

Muraki

(2011) Twitter for crisis communication: lessons learned from Japan’s tsunami disaster. International Journal of Web Based Communities 7: 392–402.

Altman

(1973) Reciprocity of interpersonal exchange. Journal for the Theory of Social Behaviour 3: 249–261.

American Red Cross (2018) Types of Emergencies. Available at: https://www.redcross.org/get-help/how-to-prepare-for-emergencies/types-of-emergencies.html

Mendiola-Smith

(2018) Mapping communication infrastructure theory onto Twitter: network integration and neighborhood storytelling. International Journal of Communication 12: 4292–4312.

Ball-Rokeach

DeFleur

(1976) A dependency model of mass-media effects. Communication Research 3: 3–21.

Ball-Rokeach

Jung

(2009) The evolution of media system dependency theory. In: Nabi

Oliver

(eds) The SAGE Handbook of Media Processes and Effects. Thousand Oaks, CA: SAGE, pp. 531–544.

Ball-Rokeach

Gonzalez

Son

, et al. (2012) Understanding individuals in the context of their environment: communication ecology as a concept and method. In: Annual conference of the International Communication Association, Phoenix, AZ, 24–28 May.

Ball-Rokeach

Kim

Y-C

Matei

(2001) Storytelling neighborhood: paths to belonging in diverse urban environments. Communication Research 28: 392–428.

Barak

Gluck-Ofri

(2007) Degree and reciprocity of self-disclosure in online forums. Cyberpsychology & Behavior 10: 407–417.

10.

boyd

Golder

Lotan

(2010) Tweet, tweet, retweet: conversational aspects of retweeting on twitter. In: Proceedings of the 43rd annual Hawaii international conference on system sciences, Honolulu, HI, 5–8 January.

11.

Burleson

Goldsmith

(1998) How the comforting process works: alleviating emotional distress through conversationally induced reappraisals. In: Andersen

Guerrero

(eds) Handbook of Communication and Emotion. San Diego, CA: Academic Press, pp. 245–280.

12.

Burt

(2004) Structural holes and good ideas. American Journal of Sociology 110: 349–399.

13.

Cha

Haddai

Benevenuto

, et al. (2010) Measuring user influence in Twitter: the million follower fallacy. In: Proceedings of the 4th international AAAI conference on weblogs and social media, Washington, DC, 23–26 May.

14.

Counts

Fisher

(2011) Taking it all in? Visual attention in microblog consumption. In: Proceedings of the fifth international AAAI conference on weblogs and social media, Barcelona, 17–21 July.

15.

Dutta-Bergman

(2006) Community participation and Internet use after September 11: complementarity in channel consumption. Journal of Computer-Mediated Communication 11: 469–484.

16.

Friedkin

(2001) Norm formation in social influence networks. Social Networks 23: 167–189.

17.

Genes

Chary

Chason

(2014) Analysis of twitter users’ sharing of official New York storm response messages. Medicine 2.0: 3: e1.

18.

Gil-Lopez

Shen

Benefield

, et al. (2018) One size fits all: context collapse, self-presentation strategies and language styles on Facebook. Journal of Computer-Mediated Communication 23: 127–145.

19.

Hawaii News Now (2018) Governor Pledges Changes after False Alert about Inbound Missile. Available at: http://www.hawaiinewsnow.com/story/37259684/ballistic-missile-threat-alert-sent-to-hawaii-phones-was-a-mistake/

20.

Heinzelman

Waters

(2010) Crowdsourcing Crisis Information in Disaster-Affected Haiti. Available at: https://www.preparecenter.org/sites/default/files/crowdsourcing_crisis_information_in_disaster-affected_haiti.pdf

21.

Heverin

Zach

(2010) Microblogging for crisis communication: examination of Twitter use in response to a 2009 violent crisis in the Seattle-Tacoma, Washington area. In: Proceedings of the 7th international ISCRAM conference, Seattle, WA, 2–5 May.

22.

Kim

Y-C

Ball-Rokeach

(2006a) Civic engagement from a communication infrastructure perspective. Communication Theory 16: 173–197.

23.

Kim

Y-C

Ball-Rokeach

(2006b) Community storytelling network, neighborhood context, and civic engagement: a multilevel approach. Human Communication Research 32: 411–439.

24.

Kim

Y-C

Jung

J-Y

(2017) SNS dependency and interpersonal storytelling: an extension of media system dependency theory. New Media & Society 19: 1458–1475.

25.

Lin

Tov

Qiu

(2014) Emotional disclosure on social networking sites: the role of network structure and psychological needs. Computers in Human Behavior 41: 342–350.

26.

List of Places in Hawaii (2018). Available at: https://en.wikipedia.org/wiki/List_of_places_in_Hawaii

27.

Matsaganis

Katz

Ball-Rokeach

(2010) Understanding Ethnic Media: Producers, Consumers, and Societies. Thousand Oaks, CA: SAGE.

28.

Mendoza

Poblete

(2010) Twitter under crisis: can we trust what we RT? In: Proceedings of the 1st workshop on social media analytics, Washington, DC, 25–28 July.

29.

Ognyanova

Ball-Rokeach

(2015) Political efficacy on the Internet: a media system dependency approach. In: Robinson

Cotten

Schulz

(eds) Studies in Media and Communications, vol. 9. Bingley: Emerald Group Publishing, pp. 3–27.

30.

Ognyanova

Jung

(2017) Digital connections: tracing the evolving role of technology in local storytelling networks. In: Kim

Y-C

Matsaganis

Wilkin

, et al. (eds) The Communication Ecology of 21st Century Urban Communities. New York: Peter Lang, pp. 146–164.

31.

Ognyanova

Chen

N-TN

Ball-Rokeach

, et al. (2013) Online participation in a community context: civic engagement and connections to local communication resources. International Journal of Communication 7: 2433–2456.

32.

Pew Research Center (2017) About 6 in 10 Young Adults in U.S. Primarily Use Online Streaming to Watch TV. Available at: http://www.pewresearch.org/fact-tank/2017/09/13/about-6-in-10-young-adults-in-u-s-primarily-use-online-streaming-to-watch-tv/

33.

Piolat

Olive

Kellogg

(2005) Cognitive effort during note taking. Applied Cognitive Psychology 19: 291–312.

34.

Wang

(2009) Online community response to major disaster: a study of Tianya Forum in the 2008 Sichuan earthquake. In: Proceedings of the 42nd Hawaii international conference on system sciences, Waikoloa, HI, 5–8 January.

35.

Rimé

(2009) Emotion elicits the social sharing of emotion: theory and empirical review. Emotion Review 1: 60–85.

36.

Rimé

Finkenauer

Luminet

, et al. (1998) Social sharing of emotion: new evidence and new questions. European Review of Social Psychology 9: 145–189.

37.

Rui

Stefanone

(2013) Strategic image management online: self-presentation, self-esteem, and social network perspectives. Information Communication Society 16: 1286–1305.

38.

Scherer

(2005) What are emotions? And how can they be measured? Social Science Information 44: 695–729.

39.

Segal

Tucker

Coolidge

(2009) A comparison of positive versus negative emotional expression in a written disclosure study among distressed students. Journal of Aggression, Maltreatment & Trauma 18: 367–381.

40.

Smith

(2013) Civic Engagement in the Digital Age. Pew Research Center. Available at http://www.ca-ilg.org/sites/main/files/file-attachments/pew_civicengagementinthedigitalage.pdf

41.

Smith

Anderson

(2018) Use of Different Online Platforms by Demographic Groups. Pew Research Center. Available at: http://www.pewinternet.org/2018/03/01/social-media-use-2018-appendix-a-detailed-table/

42.

Starbird

Palen

(2010) Pass it on? Retweeting in mass emergency. In: Proceedings of the 7th international ISCRAM conference, Seattle, WA, 2–5 May.

43.

Tateno

Yokoyama

(2013) Public anxiety, trust, and the role of mediators in communicating risk of exposure to low dose radiation after the Fukushima Daiichi Nuclear Plant explosion. Journal of Science Communication 12: 1–22.

44.

US Department of Homeland Security (2013) Innovative Uses of Social Media in Emergency Management. Available at: https://www.dhs.gov/sites/default/files/publications/Social-Media-EM_0913-508_0.pdf

45.

Van Gorp

Pogrebnyakov

Maldonado

(2015) Just keep tweeting: emergency responder’s social media use before and during Emergencies. ECIS completed research papers 191. Available at: https://research-api.cbs.dk/ws/portalfiles/portal/45521419/nicolai_pogrebnyakov_just_keep_tweeting_publishersversion.pdf

46.

Veil

Buehner

Palenchar

(2011) A work-in-process literature review: incorporating social media in risk and crisis communication. Journal of Contingencies and Crisis Management 19: 110–122.

47.

Vieweg

Hughes

Starbird

, et al. (2010) Microblogging during two natural hazards events. Proceedings of the 28th international conference on human factors in computing systems, Atlanta, GA, 10–15 April, pp. 1079–1088. New York: ACM Press.

48.

Wang

Y-C

Kraut

Levine

(2015) Eliciting and receiving online support: using computer-aided content analysis to examine the dynamics of online social support. Journal of Medical Internet Research 17: e99.

49.

Wilkin

(2013) Exploring the potential of communication infrastructure theory for informing efforts to reduce health disparities. Journal of Communication 63: 181–200.