1081046 research-article2022 JMQXXX10.1177/10776990221081046Journalism & Mass Communication QuarterlyLee et al. Original Article Journalism & Mass Communication Quarterly Visually Framing Disasters: 1­–25 © 2022 AEJMC Article reuse guidelines: Humanitarian Aid sagepub.com/journals-permissions DOI: 10.1177/10776990221081046 https://doi.org/10.1177/10776990221081046 Organizations’ Use of http://journals.sagepub.com/home/jmq Visuals on Social Media Sun Young Lee1 , JungKyu Rhys Lim1 , and Duli Shi1 Abstract The present study seeks to systematically describe how humanitarian aid organizations use visuals in their natural disaster-related social media messages and to analyze their effects on social media engagement. Using Rodriguez and Dimitrova’s (2011) four levels of visual framing, we performed a content analysis of 810 tweets from 38 aid organizations. The results showed that, overall, the organizations’ visuals had an emphasis on victims and on disaster relief efforts. The most effective types of visual framing, however, were not those the aid organizations most commonly used. We discuss the theoretical and practical implications. Keywords humanitarian aid organization, visual framing, natural disaster, social media engagement, Twitter ad creative image Natural disasters cause deaths and economic losses on a vast scale globally, killing 1.3 million people and impacting an additional 4.4 billion between 1998 and 2017 (Wallemacq & House, 2018), numbers that will only increase with increasing urban- ization and climate change (Brecht et al., 2013). Humanitarian aid organizations con- duct relief efforts during all phases of natural disasters—mitigation, preparedness, response, and recovery—focusing on areas that exceed national authorities’ relief capacity. Such disasters’ impacts are not limited to any one country and are large in 1 University of Maryland, College Park, USA Corresponding Author: Sun Young Lee, Assistant Professor, Department of Communication, University of Maryland, College Park, 4300 Chapel Lane, 2130 Skinner Building, College Park, MD 20742, USA. Email: [email protected] 2 Journalism & Mass Communication Quarterly 00(0) scale, necessitating international cooperation and long-term support. Communication is one of the many crucial challenges (e.g., logistics, supply chain management, and human resources) that aid organizations face. Disaster news coverage is transient (CARE International, 2018; Fisher et al., 2018; Houston et al., 2012) and the public’s attention span for disaster news is limited (Thrall et al., 2014), complicating efforts to connect the giving public with those currently suffering. We explored humanitarian aid organizations’ communication strategies, focusing on their visual strategies on social media. Aid organizations have increasingly used visuals to engage the public with their content and to accomplish other communication goals (PRovoke, 2017). Communicating effectively and engaging the public are cru- cial to fulfilling aid organizations’ missions, such as generating collective action, pol- icy support, fundraising, and organizational credibility. Social media are excellent tools for such purposes due to their interactivity, affordability, and availability (S. Lang, 2013; Seo & Vu, 2020). Few studies, however, have examined how aid organi- zations can best use social media strategically, and even fewer, how they can use visu- als on social media effectively (Schlimbach, 2013). This study’s purpose is twofold. First, it aims to provide a comprehensive picture of how humanitarian aid organizations use visuals in natural disaster-related messages on social media. Using Rodriguez and Dimitrova’s (2011) four-tiered model of visual framing, the study identifies visual frames at four different levels and analyzes them systematically. Second, it examines the relationships between various types of visual framing strategies and the public’s social media responses. Scholars have called for more research on how transnational organizations can effectively use social media (Seo & Vu, 2020; Thrall et al., 2014) and employ visuals in their crisis communication (Coombs & Holladay, 2011; Guidry et al., 2017), and this study is, in part, a response to that research agenda. The results will also offer practitioners guidance on the strate- gic use of visuals. Literature Review Organizations’ Use of Visuals on Social Media Visuals, as one of “the first items to catch the reader’s eye” (Rogers & Thorson, 2000, p. 8), give the first impression of a story, even for those who do not read the accompa- nying text. Telling a story with compelling visuals cuts through the clutter on social media (Walter & Gioglio, 2014), attracts attention, and conveys a story quickly, mak- ing viewers more receptive to reading the accompanying text (Brubaker & Wilson, 2018; Chung & Lee, 2019; Powell et al., 2015; Walter & Gioglio, 2014). Thus, visuals often accompany organizations’ social media messages; indeed, 83.18% of the 100 best global brands’ Facebook messages in 2013 and 79.11% in 2014 contained images (Brubaker & Wilson, 2018). Visuals can be particularly significant in humanitarian aid organizations’ social media messages. First, in the deluge of social media messages, a natural disaster is not a top-of-mind issue once it is no longer receiving media coverage. Visuals, however, Lee et al. 3 can grab people’s attention when they are quickly scrolling through social media mes- sages (Chung & Lee, 2019). Second, most people learn about disasters in other coun- tries through the media (Chouliaraki, 2006; Joye, 2009). For most people in developed countries, Joye (2009, 2010) argued, disasters are “a priori foreign news and distant suffering” (Joye, 2010, p. 255). Visuals can convey a situation more vividly than text, more effectively trigger a sense of immediacy and emotion, and surmount linguistic barriers due to their analogical quality (Borah, 2009; Rodriguez & Dimitrova, 2011). Nevertheless, how aid organizations use visuals in social media messages and the effectiveness of those visuals is poorly understood. Visual Framing of Natural Disasters Framing theory explains that how information is presented can systematically affect an audience’s interpretation and evaluation of it: “frames call attention to some aspects of reality while obscuring other elements, which might lead audiences to have differ- ent reactions” (Entman, 1993, p. 55). Geise (2017, p. 1) defined visual framing as the process of selecting some aspects of a perceived reality, highlighting them above others by means of visual communication, and making them salient, meaningful, and memorable, so that certain attributions, interpretations, or evaluations of the issue or item described are visually promoted. In the context of natural disasters, Borah (2009) compared the salient frames in the first week’s coverage of the Indian Ocean tsunami and of Hurricane Katrina in The New York Times and The Washington Post and found that the two most frequently used visual frames were lives saved (36.2%) and pragmatic (35.8%). Coverage of the two disasters differed, however—the tsunami coverage showed more death and emotional images, whereas the Katrina coverage showed more relief work and survivors. Similarly, to compare the wire services’ effect on the gatekeeping process of photos in newspapers, Fahmy et al. (2007) analyzed front-page photos in U.S. newspapers and photos distributed by two wire services in covering Hurricane Katrina and found that the depiction of flood victims was the most common, followed by the portrayal of the suffering of non-white citizens and emotion. Focusing on Twitter, Bica et al. (2017) analyzed the visual representation of two Nepal earthquakes in April and May 2015 with geotagged image tweets, finding that visuals in global tweets focused on recovery and relief work, whereas those in local tweets concentrated on people’s suffering and significant damage. Previous studies of disaster-related visuals, however, have several limitations. First, few studies have analyzed visual framing in the context of natural disasters, and most of those only examined news media coverage (Borah, 2009; Fahmy et al., 2007). The spectrum of visual framing in social media messages may be broader than that used in news media coverage, as the media usually cover stories during and immediately after a disaster, whereas aid organizations’ messages continue later in the aftermath. Furthermore, previous studies have analyzed only a few aspects of visual framing. 4 Journalism & Mass Communication Quarterly 00(0) To examine visual content systematically, we adopted Rodriguez and Dimitrova’s (2011) four levels of visual framing—the denotative, the stylistic-semiotic, the con- notative, and the ideological—where meaning progresses from the more explicit to the more implicit. First, at the denotative level, visual framing is essentially who or what is depicted, whether as discrete elements or as themes formed by combining and cat- egorizing those elements. The embedded meaning of the visuals is of less concern than what an audience receives by firsthand visual sensations. Second, the stylistic-semiotic level focuses on the stylistic conventions and how the elements or themes depicted generate social meanings. These conventions can derive from the composition of design elements within a visual, such as the social distance resulting from the camera position or angle, the prominence of an element, the pictorial expression, or the actions or poses of subjects. For example, close-up shots can convey intimacy, medium shots can signify personal relationships, and full shots can show con- text, scope, and public distance (Berger, 1991). A camera angle looking up can signify the dominant position of the subject, whereas one looking down implies the submissive position of the subject (Gale & Lewis, 2019; Hardin et al., 2002). The stylistic-semiotic level can also be deployed by editorial decisions about the location or size of a visual within media and signifies the salience of the subjects or objects in the visual. Third, at the connotative level, the subjects, objects, or themes denoted in a visual are considered as a sign or symbol of an abstract concept, idea, or meaning, whether portrayed through an abstract shape or object or figurative symbols (e.g., emblematic persons and places) (van Leeuwen, 2001), or through a visual metaphor, “a concrete image that bears some analogy to the concept” (Rodriguez & Dimitrova, 2011, p. 57). This level of visual framing is cultural or context-bound, as meanings are socially constructed, and involves moving beyond what is depicted, dwelling on the meaning, and interpreting the prominence of denotations in a visual. Fourth, at the ideological level, visuals can reflect the ideology and represent the attitudes of a nation, period, class, religion, or philosophical orientation (Panofsky, 1970). If visuals at the connotative level reify abstract ideas or meanings, the ideologi- cal level of visual framing delves into what drives those ideas or meanings. The pri- mary interest is to analyze the rationale behind the visuals (e.g., why the particular visuals were used, what interests are served), requiring a coherent interpretation of symbols and stylistic characteristics and of the larger socio-political, cultural, and reli- gious contexts. These four levels of visual framing have been applied in various contexts, such as climate change (Rebich-Hespanha et al., 2015; Wozniak et al., 2015), military con- flicts (Parry, 2011), and refugee crises (Hellmueller & Zhang, 2019; Zhang & Hellmueller, 2017). In this study, we apply the four-level model of visual framing to the natural disaster context, posing the following research questions: RQ1: At the denotative level, what elements were depicted in the visuals of aid organizations’ natural disaster-related messages? RQ2: At the stylistic-semiotic level, what camera positions and angles were used in the visuals of aid organizations’ natural disaster-related messages? Lee et al. 5 RQ3: At the connotative level, what were the dominant visual frames in the visuals of aid organizations’ natural disaster-related messages? RQ4: At the ideological level, how do the purposes of messages differ by the domi- nant visual frame in aid organizations’ natural disaster-related messages? Visual Framing and Social Media Engagement Social media engagement is a multidimensional concept encompassing the public’s interactions with social media content at the cognitive, affective, and behavioral lev- els. Some scholars have focused on the cognitive and affective aspects of engagement, defining it as “the state of cognitive and emotional absorption in the use of social media tools” (Smith & Gallicano, 2015, p. 83); others (Javornik & Mandelli, 2012; Jiang et al., 2016) have viewed engagement as a behavioral process constituting involvement, interaction, intimacy, and influence. Visuals appear to be highly effective in engaging the public on social media. Several studies have found Facebook postings with photos or graphics to be positively associ- ated with higher numbers of likes, shares, and/or comments (Abitbol & Lee, 2017; Brubaker & Wilson, 2018; Kim & Yang, 2017; Saxton & Waters, 2014). Furthermore, not only visual components’ presence, but also their framing and quality matter: Li and Xie (2020) discovered that, in tweets mentioning major airlines and SUV brands, images having human faces, professionally taken pictures, high-quality images, and relevant image content positively affected the number of likes and retweets. Ordenes et al. (2019) found that, compared with information-oriented images, action-oriented images facilitated more conversation, leading to more shares on social media. In this study, we postulate that visuals with human-interest frames—frames “por- traying one or more specific persons who are personally involved with [an] issue” (Boukes et al., 2015, p. 122)—will elicit more engagement on social media. According to exemplification theory, visuals with a human-interest frame trigger the heuristic processing of information by making the distant information more relatable, thus elic- iting emotional responses (Zillmann & Brosius, 2000). Empirical studies have demon- strated the power of visuals with human-interest frames (Boukes et al., 2015; Brantner et al., 2011; Hong, 2013). In the context of the Gaza conflict, for example, Brantner et al. (2011) conducted an experiment with human-interest framing versus visual polit- ical framing versus no visual framing and found that human-interest framing, focusing on individual victims and civilians, elicited stronger emotional effects and more posi- tive evaluations of news stories. We also posit that visuals portraying negative facial expressions of victims will generate more engagement on social media than will other types of visuals. Zillmann and Brosius (2000) proposed that the effects of exemplars—in our study context, per- sons in a human-interest frame in a visual—will be stronger when the exemplars engage the emotions more. Facial expressions are one important cue through which viewers can discern people’s emotional state (Hong, 2013). Moreover, studies have shown the effectiveness of negative visuals. The limited capacity model of motivated mediated message processing (LC4MP) suggests that, at the moderate level of arousal 6 Journalism & Mass Communication Quarterly 00(0) content, negative visuals are more effective than positive ones, especially negative visuals with higher arousal levels (A. Lang, 2006). Based on the LC4MP, Chung and Lee (2019) found that arousing negative images worked best in eliciting attributions of public-serving corporate social responsibility (CSR) motives, favorable attitude toward companies leading a CSR initiative, and behavioral intentions toward the com- panies’ products or CSR initiatives. In the context of the BP oil spill, Miller and LaPoe (2016) found that people remembered visuals triggering negative emotions the most (e.g., oil-soaked animals). Furthermore, we predict that visual frames showing organizations’ identities will help in engaging the public. According to brand management research, brand logos and names facilitate brand familiarity and recognizability and can trigger the brand’s existing associations and related beliefs, increasing favorable attitudes and compli- ance with the brand’s recommendations (Foroudi, 2019; van Grinsven & Das, 2016). The link between a logo and organizational identity is especially noticeable for value- driven organizations (Erjansola et al., 2021). Visually displaying an aid organization’s identity, therefore, may evoke its mission and values, legitimizing its work or generat- ing confidence in it, leading to more social media engagement. Therefore, we propose the following hypotheses: H1: In humanitarian aid organizations’ social media messages, (a) visuals portray- ing human-interest frames, (b) those portraying the negative facial expressions of victims, and (c) those indicating organizations’ identities generate higher levels of social media engagement than do other types of visuals. In addition, we propose a research question about the effects of other types of visual framing on social media engagement. Research predicting the relationships is lacking, but other types of visual framing strategies identified in answering RQs 1 to 4 are worth investigating to deepen our understanding of a wider spectrum of visual strate- gies and to stimulate future theoretical developments. RQ5: Do any other types of visual framing generate higher levels of social media engagement? Method Study Organizations The population of organizations was humanitarian aid organizations. We compiled comprehensive lists of three types of humanitarian aid organizations—the United Nations system, international non-governmental organizations (NGOs), and national governmental organizations. We included all the organizations from Stoddard (2003), which listed international NGOs, and from the World Confederation of Physical Therapy (2019), which covered all three types of organizations. We used three other sources to supplement the list of national governmental organizations (Jennings, 2017; Lee et al. 7 the Government of India, 2021; the Organisation for Economic Co-operation and Development [OECD], 2020), ultimately identifying 47 humanitarian aid organizations. Data Collection First, we identified tweets from the study organizations containing natural disaster- related messages from July 1, 2018, to June 30, 2019. The timeframe captured one full year, as some natural disasters are seasonal. We chose Twitter due to its popularity during crises and emergencies (Spence et al., 2015). When an organization had mul- tiple accounts, we used its global account. If an organization had multiple country- specific accounts, we selected the account with the greatest number of followers. We also limited the language to English. To retrieve the data, we collected the tweets with the Twitter application program- ming interface (API) using the “GET statuses/user_timeline” command from each organization’s account. When we reached the maximum downloadable number of tweets (3,200 per account; Twitter, 2020), we manually downloaded tweets via Twitter Advanced Search. Then, using a machine-aided process, we filtered for tweets con- taining a visual component and any of the 84 natural disaster-related keywords we had identified based on the names of natural disasters during the timeframe and the types of disaster in the Emergency Events Database (Centre for Research on the Epidemiology of Disasters, 2020). To eliminate possible confounding effects from different modali- ties (e.g., Abitbol & Lee, 2017; Brubaker & Wilson, 2018), we limited our visuals to still images of actual objects, leaving 810 tweets from 38 organizations. Coding Variables First, we coded each tweet’s background information (e.g., the posting date, type of disaster, disaster name, and country affected by the disaster) and then the main vari- ables, using Rodriguez and Dimitrova’s (2011) four levels of visual framing. Denotative level (RQ1).  We examined the portrayal of three elements—victims, other subjects, and organizations—commonly presented in visuals of natural disaster-related messages (Bica et al., 2017; Borah, 2009; Fahmy et al., 2007; Rebich-Hespanha et al., 2015; Wozniak et al., 2015). Portrayal of victims.  We defined victims as people “directly affected by disastrous events” (Ladrido-Ignacio & Perlas, 1995, p. 9). We coded whether a visual contained a victim or not and the number of victims, if present, using the following catego- ries: no victim in the image, one individual, a small group of two to three victims, a medium group of four to eight victims, a focus on a victim(s) with a group of people in the background, a large group of nine or more, or other. Focusing on a smaller number of victims was more likely to emphasize individuals’ stories or a human- interest framing. 8 Journalism & Mass Communication Quarterly 00(0) If one or more victims were portrayed, we also coded their characteristics by the presence (1) or absence (0) of each category in the following variables: age group (baby/toddler, child, adult, elderly, or not sure); gender (male, female, or not sure); family; and victims’ facial expression in terms of valence and intensity (not recogniz- able/ambiguous, high-arousal positive, low-arousal positive, neutral, low-arousal neg- ative, or high-arousal negative) (Berger, 1991; Hellmueller & Zhang, 2019; Zhang & Hellmueller, 2017). The unit of coding was a tweet, which allowed for the presence of multiple categories when more than one victim was portrayed. Portrayal of other subjects.  We coded the presence (1) or absence (0) of different types of actors other than victims: volunteer or responder, professional, or other. Again, as the unit of coding was a tweet, the presence of multiple categories was permitted. Visibility of the organization.  We coded the presence (1) or absence (0) of the identity of the aid organization or its partner organization(s) through a logo or a name. Stylistic-Semiotic Level (RQ2). We coded the camera shot position and camera angle, adapting previous studies (Berger, 1991; Hellmueller & Zhang, 2019; Zhang & Hell- mueller, 2017 see the images in the Online Material section on the JMCQ website). We created three categories for the camera shot position—close-up, medium shot, and long shot—based on the distance between the camera and the subjects or objects focused on in the frame. We coded the camera angle into three categories depending on the location of the camera: looking down (the camera being higher than the subject or object focused on), looking straight on (level), and looking up. The connotative level (RQ3).  We defined the visual framing at the connotative level as a dominant visual frame—one that emphasizes a certain part of the issue, channels the meaning, and influences the definition and interpretation of the problem (Brantner et al., 2011). Examining this involves a latent level of analysis, as it derives from the complex of denotative elements and stylistic factors and highlights the most represen- tative, core idea attached to a visual. We first deductively developed categories based on the literature (Borah, 2009; Brantner et al., 2011; Fahmy et al., 2007; Hellmueller & Zhang, 2019; Houston et al., 2012; Zhang & Hellmueller, 2017). Then, we inductively added categories while itera- tively reviewing visuals used in tweets from the study organizations, generating nine categories (see Table 1 for the descriptions and the Online Material section on the JMCQ website for examples). All categories were exclusively coded, except the cate- gory of disaster relief efforts. The actions of responders to a disaster were often emphasized equally with victims in distress or other categories. The ideological level (RQ4).  To analyze the ideological level of visual framing, we coded the purpose of the message by examining the accompanying text and compared it with the visual frames coded at the connotative level. The text often illuminates the purpose of a tweet and offers clues to the rationale for using a visual frame. Based on the Lee et al. 9 Table 1.  Description of Dominant Visual Frames and Purpose of Text (RQ3 & RQ4). Variable Category Description Dominant visual Victims in need or Highlights victims in need or distress. The frame distress human dimension is highlighted, showing those grieving, suffering, or in pain.   Victims recovering Highlights an individual or group who has fully recovered or is in the process of recovering from the disaster.   Disaster preparation Features people taking action to prevent the efforts damage from a disaster.   Normal scenery Features overall landscape of normal scenery. Shows what it looked like before a disaster or after recovery from a disaster.   Disaster damage Shows a scene of destruction from the scenery disaster such as the physical damage to buildings, homes, or the area.   Disaster relief efforts Highlights the actions of people aiding disaster victims or preparing or delivering disaster aid materials.   Community resilience Features a new outlook or growth or a return to the normal life of the community.   Symbolic object Features a symbolic object(s); not a snapshot of daily life or reality.   Meeting Conference, panel, summit, or meetings discussing disaster relief efforts or sharing knowledge about a natural disaster. Purpose of text Organizational credit Showcases or highlights the organization’s efforts and contributions.   Delivering disaster Shares information about a disaster or an news area/people impacted by the disaster.   Disaster resilience Showcases or celebrates the survivors of or recovery from a disaster.   Communicative Urges the public to like/share/comment on action the message or to seek more information.   A call to action Asks the public to donate, volunteer, or provide other help.   Educational Gives the public knowledge about a natural information disaster from research, industry reports, or experts. literature on nonprofits’ social media communications (Powers, 2018; Saxton & Waters, 2014; Seo & Vu, 2020) and on tweets from the study organizations, we devel- oped six categories (see Table 1 for the descriptions). As a text message can have more than one purpose, we coded each of the categories as presence (1) or absence (0) and then compared the purpose(s) of the messages across the dominant visual frames. 10 Journalism & Mass Communication Quarterly 00(0) Social media engagement (H1 and RQ5).  H1a predicted the significant effects of human- interest frames, defined as visuals portraying a victim, volunteers, or responder, or professional at the denotative level (see Table 2) and the dominant visual frames of victims in need or distress, victims recovering, and disaster relief efforts at the con- notative level (see Table 4). H1b postulated the effects of portraying the negative facial expressions of victims using the categories of low-arousal and high-arousal negative facial expressions (see Table 2). H1c posited the effects of the visibility of organiza- tions’ identities (see visibility of the organization in Table 2). We measured social media engagement as the number of likes, retweets, and replies, which, as built-in-features on Twitter, are widely used indicators of social media engagement, especially in content analysis (e.g., Abitbol & Lee, 2017; Brubaker & Wilson, 2018; Kim & Yang, 2017). These three manifest indicators of engagement represent the level of engagement with each message, making them appropriate mea- sures for testing the effects of message strategies. We retrieved the number of likes and retweets using the Twitter API and manually coded the number of replies. Intercoder Reliability All authors coded the tweets with one of them serving as the main coder, training the other coders. Each coder participated in six or seven rounds of training sessions by independently coding tweets outside those used in this study. We calculated intercoder reliability scores by all coders’ independently coding a subset of the actual data—81 tweets (10% of the total number of tweets) (Neuendorf, 2010; Riffe et al., 2019). The simple agreement ranged from 95.89% to 100% and Krippendorff’s alpha ranged from .84 to 1, both of which are above the cutoff points (Krippendorff, 2004; see Tables 2–4). Results Among the 38 organizations, the International Federation of the Red Cross and Red Crescent Societies (IFRC) issued the most tweets (n = 142, 17.53%), followed by the World Food Program (WFP) (n = 56, 6.91%), and the UN Development Program (UNDP) (n = 53, 6.54%). The average number of likes was 63.33 (SD = 115.65), ranging from 0 to 1,362. The average number of retweets was 30 (SD = 57.3), ranging from 0 to 928. The average number of replies was 171 (SD = 6.86), ranging from 0 to 171. The Twitter accounts varied widely in the number of followers, ranging from 445 to 7,590,000. To answer RQs 1 to 3, we ran descriptive statistics for each category within each variable. The Denotative Level (RQ1) The majority of tweets portrayed victims (n = 453, 55.93%); among these, the most popular types of framing were a focus on an individual (n = 165, 20.37%) or a small group of victims (n = 146, 18.02%). In terms of victims’ characteristics, most were Lee et al. 11 Table 2.  Summary of Visual Framing at the Denotative Level (RQ1). Kripendorff’s Variable Category n % alpha (α) Portrayal of victims (N = 810) (H1a)     No victim 357 44.07 .95   One individual 165 20.37     Small group 146 18.02     Medium group 34 4.20     Focused on individual(s) 32 3.95     Large group 30 3.70     Other 46 5.68   Characteristics of victims (N = 416)    Age Baby/toddler 96 23.08 .97 Child 162 38.94 .98 Adult 286 68.75 .95 Elderly 28 6.73 .86 Not sure 1 0.24 NA  Gender Male 176 42.31 .94 Female 274 65.87 .95 Not sure 126 30.29 .84  Family Presence 135 32.45 .95   Facial expression High-arousal positive 37 8.89 .95 Low-arousal positive 81 19.47 .89 Neutral 233 56.01 .91 Low-arousal negative (H1b) 34 8.17 1 High-arousal negative (H1b) 1 0.24 NA Not recognizable/ambiguous 61 14.66 .87 Other subjects (N = 810)     No other subject 526 64.94 1   Volunteer or responder (H1a) 221 27.28 1   Professional (H1a) 30 3.70 1   Other 43 5.31   Visibility of the organization (N = 810) (H1c)     Presence 259 31.98 .91 Note. The coders chose one of the seven categories in the variable of the portrayal of a victim(s), so there is only one set of intercoder reliability scores. The victims’ characteristics were only coded when a tweet portrayed at least one victim and victims’ characteristics were recognizable. We used Krippendorff’s alpha (α) to assess intercoder reliability across all coders of subsamples (n = 81, 10%) of actual coding materials; the visual framing strategies referred to in H1 are indicated (i.e., H1a, H1b, and H1c). adult (n = 486, 68.75%), but child (n = 162, 38.94%) and baby/toddler (n = 96, 23.08%) were also fairly common; visuals depicted more female (n = 274, 65.87%) than male (n = 176, 42.31%) subjects. About a third of the tweets (32.45%, n = 135) 12 Journalism & Mass Communication Quarterly 00(0) Table 3.  Summary of Visual Framing at the Stylistic-Semiotic Level (RQ2). Variable Category n % Kripendorff’s alpha (α) Camera shot position Close-up 107 13.21 .91   Medium shot 610 75.31     Long shot 93 11.48   Camera angle Looking down 91 11.23 .97   Looking straight 682 84.20     Looking up 37 4.57   Note. N = 810. The coders chose one of the three categories for the camera shot position and the camera angle, respectively, so there is only one set of intercoder reliability scores for each variable. We used Krippendorff’s alpha (α) to assess intercoder reliability across all coders of subsamples (n = 81, 10%) of actual coding materials. depicted a family. The majority of the victims’ facial expressions were neutral (n = 233, 56.01%), but victims more often had positive facial expressions (n = 118, 28.37%, combining high and low arousal levels) than negative ones (n = 35, 8.41%). The majority of visuals (n = 526, 64.94%) did not contain other subjects, but when shown, the most frequently portrayed other subjects were volunteers or responders (n = 221, 27.28%). Finally, 259 (31.98%) tweets presented the identity of an aid organi- zation or its partner organization(s) (see Table 2). The Stylistic-Semiotic Level (RQ2) A medium shot was the most frequent camera position (n = 610, 75.31%), followed by a close-up (n = 107, 13.21%) and a long shot (n = 93, 11.48%). The majority of visuals had a camera angle looking straight on (n = 682, 84.2%), followed by a camera angle looking down (n = 91, 11.23%) and a camera angle looking up (n = 37, 4.57%) (see Table 3). The Connotative Level (RQ3) The two most dominant visual frames were victims in need or distress (n = 389, 48.02%) and disaster relief efforts (n = 203, 25.06%), which together are connected to the aid organizations’ mission. Among these, 63 tweets (7.78%) emphasized visual frames both of victims in need or distress and of disaster relief efforts (see Table 4). The Ideological Level (RQ4) RQ4 asked how the purpose of the messages differed by dominant visual frame. We first examined the purpose of the message inferred from the accompanying text. The two most frequently identified purposes were to deliver disaster news (n = 495, 61.11%)—implying that aid organizations may function as a news channel even when the news media are no longer covering the issue—and to earn credit for and establish the legitimacy of their work (n = 387, or 47.78%; see Table 4). Lee et al. 13 Table 4.  Summary of Visual Framing at the Connotative Level (RQ3) and Purpose of Message (RQ4). Variable Category n % Kripendorff’s alpha (α) Dominant visual Victims in need or distress (H1a) 389 48.02 .97 frame Victims recovering (H1a) 39 4.81 1 Disaster preparation efforts 43 5.31 1 Normal scenery 24 2.96 1 Disaster damage scenery 81 10 .97 Disaster relief efforts (H1a) 203 25.06 .98 Community resilience 8 0.99 1 Symbolic object 30 3.70 1 Meeting 17 2.10 1 Other 51 6.30   Purpose of the Organizational credit 387 47.78 .92 message Delivering disaster news 495 61.11 .98 Disaster resilience 19 2.35 .98 Communicative action 85 10.49 .92 Call to action 72 8.89 .84 Educational information 122 15.06 .94 Other 231 28.52   Note. N = 810. We used Krippendorff’s alpha (α) to assess intercoder reliability across all coders of subsamples (n = 81, 10%) of actual coding materials; the visual framing strategies referred to in H1 are indicated (i.e., H1a, H1b, and H1c). Next, we compared the purposes of messages for each of the dominant visual frames, focusing on the five most frequently used dominant visual frames and the five most common purposes of the messages (see Figure 1). Then, we ran a series of logistic regression analyses to further examine the association between the purposes of mes- sages and each of the visual frames. The purposes of messages served as independent variables and were entered all together for each of the visual frames, each frame being a binary outcome coded as presence or absence. At the ideological level, the premise is that a purpose lies behind the visual framing. The purposes of delivering disaster news (B = 1.16, SE = .17, OR = 3.20, p < .001), call to action (B = 1.13, SE = .29, OR = 3.09, p < .001), and educational information (B = 0.58, SE = .24, OR = 1.79, p = .017) were significantly associated with the use of a victims in need or distress visual frame. The purpose of delivering disaster news was also significantly associated with the use of a disaster damage scenery frame (B = 1.21, SE = .38, OR = 3.34, p < .001). The purpose of organizational credit was significantly associated with the use of a disaster relief efforts frame (B = 1.18, SE = .19, OR = 3.25, p < .001). Social Media Engagement (H1 and RQ5) To test H1a, H1b, and H1c, and to answer RQ5, we conducted a series of fixed-effects negative binomial regression analyses by regressing social media engagement on each 14 Journalism & Mass Communication Quarterly 00(0) Figure 1.  Purpose of messages within each dominant visual frame. Note. The percentage was calculated based on the total number of tweets for each of the dominant visual frames. level of visual framing. We used fixed-effects negative binomial regression analyses because the three dependent variables—the number of likes, retweets, and replies— were count variables and skewed to the right, following a Poisson distribution. Furthermore, the descriptive statistics showed overdispersion of the data, fitting a nega- tive binomial regression model. In addition, we used fixed-effects models to control for all non-observed sources of variation that were correlated with the exposure variable— the organization, in this case—and the outcome variables. We also controlled for the number of followers, as they varied significantly among the organizations, which might correlate with the number of likes, retweets, or replies (see Table 5). Human-interest frames (H1a).  For human-interest frames (H1a), we examined whether the visuals portraying a victim, volunteer or responder, or professional at the denota- tive level and the dominant frames of victims in need or distress, victims recovering, or disaster relief efforts at the connotative level generated higher social media engage- ment. None of the denotative-level variables were significantly associated with the number of likes, retweets, or replies. At the connotative level, using a disaster relief efforts visual frame (B = 0.15, SE = .07, IRR = 1.16, p = .025) was significantly associated with a higher number of likes. Thus, H1a was supported only for the disas- ter relief efforts frame on the number of likes. Victims’ negative facial expressions (H1b). For victims’ negative facial expressions (H1b), we examined the relationships between visuals portraying low- and high- arousal negative facial expressions of victims at the denotative level and social media engagement. Tweets portraying victims’ negative facial expressions were not signifi- cantly associated with a higher number of likes, retweets, or replies. Therefore, H1b was not supported.  Table 5.  The Effects of Visual Framing on Social Media Engagement (H1 and RQ5). Likes Retweets Replies Category β (SE1) IRR (SE2) β (SE1) IRR (SE2) β (SE1) IRR (SE2) Model 1: Denotative (N = 810)   Portrayal of victims (H1a) (reference group: no victim portrayed)   One individual −0.05 (0.06) 0.95 (0.06) −0.25 (0.07)** 0.78 (0.06)** −0.35 (0.13)** 0.7 (0.09)**   Small group −0.11 (0.07) 0.9 (0.06) −0.25 (0.07)** 0.78 (0.06)** −0.28 (0.13)* 0.75 (0.1)*   Medium group 0 (0.11) 1 (0.11) −0.24 (0.12) 0.79 (0.1) −0.24 (0.22) 0.79 (0.17)    Focused on individual(s) 0.02 (0.11) 1.02 (0.11) −0.04 (0.12) 0.97 (0.11) 0 (0.2) 1 (0.2)   Large group −0.01 (0.12) 0.99 (0.12) −0.25 (0.14) 0.78 (0.11) −0.26 (0.26) 0.77 (0.2)   Volunteer or responder (H1a) 0.08 (0.06) 1.08 (0.07) −0.05 (0.07) 0.95 (0.06) −0.03 (0.12) 0.97 (0.12)   Professional (H1a) −0.48 (0.15)** 0.62 (0.09)** −0.6 (0.17)*** 0.55 (0.09)*** −1.04 (0.36)** 0.35 (0.13)**   Visibility of the organization (H1c) 0.12 (0.06)* 1.13 (0.07)* 0.1 (0.06) 1.11 (0.07) 0.21 (0.12) 1.24 (0.14) Model 2: Denotative (characteristics of victims; N = 416)  Age     Baby/toddler 0.03 (0.11) 1.03 (0.11) −0.1 (0.12) 0.9 (0.11) −0.07 (0.23) 0.94 (0.21)   Child −0.06 (0.09) 0.94 (0.09) −0.09 (0.1) 0.91 (0.09) 0.15 (0.2) 1.17 (0.23)   Adult −0.2 (0.11) 0.82 (0.09) −0.18 (0.12) 0.83 (0.1) −0.06 (0.22) 0.94 (0.21)   Elderly −0.1 (0.14) 0.9 (0.12) 0.03 (0.14) 1.03 (0.15) 0.01 (0.28) 1.01 (0.28)  Gender     Male 0.05 (0.07) 1.06 (0.08) 0.06 (0.08) 1.06 (0.09) −0.05 (0.15) 0.95 (0.15)   Female 0.04 (0.08) 1.04 (0.08) −0.04 (0.08) 0.96 (0.08) 0.08 (0.17) 1.09 (0.19)  Family 0.05 (0.11) 1.05 (0.11) 0.14 (0.12) 1.15 (0.13) 0.21 (0.22) 1.24 (0.28)   Facial expression     High-arousal positive 0.42 (0.11)*** 1.52 (0.16)*** 0 (0.12) 1 (0.13) −0.07 (0.28) 0.93 (0.26)   Low-arousal positive 0.12 (0.09) 1.13 (0.1) −0.11 (0.1) 0.9 (0.09) −0.01 (0.2) 0.99 (0.19)   Neutral 0.16 (0.08)* 1.18 (0.1)* 0.01 (0.09) 1.01 (0.09) 0.12 (0.17) 1.13 (0.19)    Low-arousal negative (H1b) 0.01 (0.12) 1.01 (0.12) −0.09 (0.13) 0.91 (0.12) 0.08 (0.25) 1.08 (0.27)    High-arousal negative (H1b) 0.21 (0.52) 1.24 (0.65) −1.03 (0.81) 0.36 (0.29) −18.97 (15604.93) 0 (0) (continued) 15  16 Table 5.  (continued) Likes Retweets Replies Category β (SE1) IRR (SE2) β (SE1) IRR (SE2) β (SE1) IRR (SE2) Model 3: Stylistic-semiotic (N = 810)   Camera shot (reference group: close-up)   Medium shot −0.1 (0.06) 0.9 (0.06) 0.02 (0.07) 1.02 (0.08) 0.06 (0.13) 1.07 (0.14)   Long shot 0.04 (0.09) 1.04 (0.09) 0.29 (0.1)** 1.33 (0.13)** 0.2 (0.18) 1.23 (0.22)   Camera angle (reference group: looking down)   Looking straight −0.19 (0.07)** 0.83 (0.06)** −0.26 (0.07)** 0.77 (0.06)** −0.26 (0.14) 0.77 (0.1)   Looking up −0.37 (0.13)** 0.69 (0.09)** −0.45 (0.14)** 0.64 (0.09)** −0.54 (0.28) 0.58 (0.16) Model 4: Connotative (N = 810) Journalism & Mass Communication Quarterly 00(0)   Victims in need or distress (H1a) −0.14 (0.07) 0.06 (−1.88) 0.08 (−1.51) 0.89 (0.07) −0.19 (0.14) 0.83 (0.12)   Victims recovering (H1a) 0.04 (0.11) 1.05 (0.12) −0.1 (0.13) 0.9 (0.12) −0.03 (0.24) 0.97 (0.23)   Disaster preparation efforts −0.14 (0.12) 0.87 (0.11) −0.09 (0.13) 0.92 (0.12) −0.3 (0.25) 0.74 (0.19)   Normal scenery −0.13 (0.16) 0.88 (0.14) −0.05 (0.17) 0.95 (0.16) −0.35 (0.33) 0.71 (0.24)   Disaster damage scenery −0.05 (0.1) 0.95 (0.1) 0.25 (0.1)* 1.29 (0.14)* 0.3 (0.18) 1.34 (0.25)   Disaster relief efforts (H1a) 0.15 (0.07)* 1.16 (0.08)* 0.12 (0.07) 1.13 (0.08) 0.22 (0.13) 1.25 (0.16)   Symbolic object 0.07 (0.15) 1.07 (0.16) 0.16 (0.16) 1.18 (0.19) −0.25 (0.3) 0.78 (0.23)  Meeting −0.14 (0.19) 0.87 (0.16) −0.15 (0.21) 0.86 (0.18) −0.18 (0.38) 0.84 (0.32) Note. Each model (Model 1–Model 4) was run in a separate equation because the categories in the different levels of framing can overlap due to the hierarchical nature of the concepts. We performed two separate regression analyses at the denotative level because the victim characteristics variables were only applied to tweets containing images of a victim(s). The visual framing strategies referred to in H1 are indicated (i.e., H1a, H1b, H1c). If the coefficient (β) was negative, employing the message strategy drew even less engagement than the reference group did; therefore, even if a coefficient was statistically significant, if it was negative, the visual framing strategy was not effective at drawing more engagement. IRR = incidence-rate ratio; since the IVs are categorical variables, IRR indicates the ratio of the differences in DVs between a group and its reference group; for example, if the value of the IRR was 2, the group had twice as much in a DV as its reference group; if the value of the IRR was less than 1, the reference group had a higher number in a DV. SE1 = robust standard error of the coefficient; SE2 = robust standard error of the IRR. *p < .05. **p < .01. ***p < .001. Lee et al. 17 Visibility of the organization (H1c).  For organizational visibility (H1c), we examined the effects of visuals where an organization’s identity was visible, which was at the deno- tive level of framing, on social media engagement and found that tweets indicating organizations’ identities generated a significantly higher number of likes (B = 0.12, SE = .06, IRR = 1.13, p = .038). Thus, H1c was supported, but only for the number of likes. Other types of visual framing (RQ5).  Finally, we examined the relationship between all other types of visual framing and social media engagement (RQ5). First, contrary to H1b, visuals of victims with a high-arousal positive facial expression (B = 0.42, SE = .11, IRR = 1.52, p < .001) and with a neutral facial expression (B = 0.16, SE = .08 IRR = 1.18, p =.042) generated significantly more likes than tweets lacking those ele- ments. At the stylistic-semiotic level, the tweets having a camera angle looking down were more effective in generating likes than those looking straight on (B = −0.19, SE = .07, IRR = 0.83, p = .008) or looking up (B = −0.37, SE = .13, IRR = 0.69, p = .005) and were more effective in generating retweets than those looking straight on (B = −0.26, SE = .07, IRR = 0.77, p = .001) or looking up (B = −0.45, SE = .14, IRR = 0.64, p = .002). Moreover, long camera shots were more effective than close-ups (B = 0.29, SE = .10, IRR = 1.33, p = .004) in generating retweets. At the connotative level, the disaster damage scenery visual frame generated significantly more retweets (B = 0.25, SE = .10, IRR = 1.29, p = .015). None of the visual framing strategies elicited a significantly higher number of replies. Discussion With the continuing devastation wrought by natural disasters, aid organizations’ ongo- ing and strategic communication has become more important than ever. This study sought to examine the visual framing strategies of aid organizations and their effectiveness. The results of the content analysis of visual framing showed that, overall, aid orga- nizations frequently used human-interest frames, focusing on victims and volunteers or responders. The results provide empirical evidence for what journalists and com- munication practitioners have emphasized as effective visual frame-building strate- gies: the use of humanized and personalized images (Dhanesh & Rahman, 2021). However, although visuals from both aid organizations and the news media commonly portrayed victims (Borah, 2009; Fahmy et al., 2007; Houston et al., 2012), aid organi- zations’ frequent use of visuals portraying disaster relief efforts and volunteers or responders contrasted with the news media’s frequent use of visuals of disaster dam- age scenery. The results highlight aid organizations’ missions and communication purposes. Furthermore, this study delved into how victims were visually portrayed and other aspects of visual framing. The results showed that the portrayals of victims were somewhat stereotypical, as many of the tweets portrayed a baby (23%) or a child (39%) and a majority depicted women. However, the portrayals of victims were more 18 Journalism & Mass Communication Quarterly 00(0) nuanced emotionally, mostly showing victims with neutral facial expressions. The results reveal a gap between professional norms emphasizing the use of a visual hook—“an image that represents the narrative and grabs the audience’s attention” (Dhanesh & Rahman, 2021, p. 5)—and the actual practices. As to how the various types of visual framing impacted the public’s engagement on social media, the present results showed that the most frequently used visual framing was not the most effective one. First, among human-interest frames, victim-focused frames were prevalent both in the literature (Borah, 2009; Fahmy et al., 2007; Houston et al., 2012) and in this study. However, a responder-oriented human-interest frame was more effective, a finding that opens up unique opportunities for aid organizations. Second, although visual frames evoking positive emotions were not frequently used, the results demonstrated the value of positive emotions in a natural disaster con- text, providing empirical support for practitioners’ employing positive visual framing that reinforces hope, resilience, and courage, rather than using undignified, helpless, stereotypical portrayals of victims (Dhanesh & Rahman, 2021). The results contra- dicted our hypothesis and the LC4MP, but one possible explanation is that the effec- tiveness of the emotional valence of visual framing varies by context. Studies have found that during disasters, individuals mostly shared positive content on social media, such as sympathy, emotional support, cautions and advice, concern for the affected, and donation and volunteer opportunities (Buntain & Lim, 2018; Olteanu et al., 2015), and that hope triggered individuals to seek and process information (Yang et al., 2010, 2011) and to take protective actions (Lim et al., 2019). Although these were not visual- specific studies, the same logic might have applied in this study. In addition, Twitter’s “like” feature may have contributed to the popularity of positive frames, as, unlike Facebook, Twitter has no option to select different emotional reactions, so users are unlikely to approvingly “like” an image of suffering. Another important finding is the significant effects of the visibility of an organiza- tion’s identity. Studies have shown that due to existing familiarity with a brand, a brand’s logo or name can trigger pre-existing associations, affecting interest in and the perception of messages from the brand (Foroudi, 2019; van Grinsven & Das, 2016). In addition to testing the hypotheses, the results answering RQ5 yield additional insights. Specifically, more retweets were generated by visuals of damaged scenery using a long shot, and more likes and retweets were elicited by a downward-looking camera angle. A camera angle looking down signifies a submissive subject and a dom- inant viewer, putting the general public in a position superior to the victims (Gale & Lewis, 2019; Hardin et al., 2002). Such visuals possibly triggered sympathy or engaged the viewers more emotionally. In addition, an aerial shot allows viewers to see the overall disaster scene in a neutral way, which might have aided in the cognitive pro- cessing of information, permitting a holistic assessment of the situation. Theoretical and Practical Implications This study makes both theoretical and practical contributions. First, it contributes to the scholarship on aid organizations’ social media communication strategies. Despite Lee et al. 19 social media’s advantages in directly connecting with the public on globally relevant topics, how aid organizations use social media and how they could use them more effectively is little understood. The literature has explored these topics more broadly, such as whether these organizations use social media or not and the factors that influ- ence their social media use (e.g., Seo & Vu, 2020; Thrall et al., 2014). This study’s focus on message strategies offers a more fine-grained picture of how these organiza- tions use social media and gives insights into their strategies. Second, our study advances the research on visual framing. By applying Rodriguez and Dimitrova’s (2011) four levels of visual framing to the natural disaster context, this study offers a systematic analysis of natural disaster-related visuals and demon- strates the applicability of the theory to the study context. Several studies have exam- ined visual framing in the context of natural disasters, but their analysis of visual framing has been limited to only a few aspects of such framing (e.g., Borah, 2009; Fahmy et al., 2007). Visual framing studies have mostly focused on the news media, following in the tradition of the framing research that originated in studies of the news media (Borah, 2009; Fahmy et al., 2007; Houston et al., 2012); how other types of organizations frame visuals has been little explored (Dhanesh & Rahman, 2021; Li & Xie, 2020). This study showed that aid organizations’ framing strategies can exhibit more variety because they cover long-term, ongoing recovery situations, whereas news media coverage focuses on the period during or immediately after a disaster. Third, our study contributes to the growing literature on the effectiveness of visual strategies. Empirical studies on visual framing are mostly descriptive in nature and studies on its effects are still few in number and lacking comprehensiveness, espe- cially in the study context. This study contributes to advancing knowledge in this underexplored area while validating exemplification theory and extending its applica- tion to a broader range of visual framing. The study also opens further avenues of research on the LC4MP, such as testing the effects of emotional visuals in a more controlled setting. In terms of practice, the study offers guidance on visual communication strategies. By looking at which types of visuals are most frequently used in aid organizations’ natural disaster-related messages, professionals can better understand organizational norms in practice. More importantly, the results showing the effects of different types of visual framing on social media engagement suggest that practitioners might benefit by employ- ing human-interest frames: victims to enable the public to connect better to distant suf- fering; victims’ positive facial expressions, aid workers, and disaster relief efforts to convey resilience and hope; organizational names and logos to legitimize disaster response efforts; and downward-looking camera shots to evoke affective engagement. Long-shot views of disaster damage might spur cognitions about the extent of the dam- age. Using visuals in conjunction with the purposes of texts will also be important. Limitations and Conclusion This study has several limitations. First, we only focused on a 1-year timeframe. The types of visual framing that we analyzed may be tied to the characteristics of the 20 Journalism & Mass Communication Quarterly 00(0) specific disasters that occurred within that period. A longitudinal study might show other trends. Second, the results cannot be taken to show causal relationships. Future research can use experiments to test the effects of visual framing in a controlled set- ting. Third, future research can compare visual framing across different types of media, organizations, and countries. Finally, the results are not generalizable to other media, organizations, or channels, as we only examined Twitter and well-known organiza- tions. Future research can cast a wider net. Despite its limitations, this study lays a foundation for future research in this little- explored area and suggests effective strategies for engaging the public with humanitar- ian aid organizations on social media. With global climate change, natural disasters will be increasingly relevant in the future. The public’s engagement with relief efforts can be enhanced by the continuing endeavors of both academics and practitioners. Acknowledgments The authors would like to thank Save the Children, the United Nations Development Programme (UNDP), and the United States Agency for International Development (USAID) for generously allowing them to use examples of their social media visuals. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. ORCID iDs Sun Young Lee https://orcid.org/0000-0002-4562-1279 JungKyu Rhys Lim https://orcid.org/0000-0002-5006-2491 Duli Shi https://orcid.org/0000-0002-9058-952X Supplemental Material Supplemental material for this article is available online. References Abitbol, A., & Lee, S. Y. (2017). Messages on CSR-dedicated Facebook pages: What works and what doesn’t. Public Relations Review, 43(4), 796–808. https://doi.org/10.1016/j. pubrev.2017.05.002 Berger, A. A. (1991). Media USA: Process and effect. Longman. Bica, M., Palen, L., & Bopp, C. (2017). Visual representations of disaster. In Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing (pp. 1262–1276). Association for Computing Machinery. https://doi. org/10.1145/2998181.2998212 Lee et al. 21 Borah, P. (2009). Comparing visual framing in newspapers: Hurricane Katrina versus tsunami. Newspaper Research Journal, 30(1), 50–57. https://doi.org/10.1177/073953290903000106 Boukes, M., Boomgaarden, H. G., Moorman, M., & de Vreese, C. H. (2015). Political news with a personal touch: How human interest framing indirectly affects policy atti- tudes. Journalism & Mass Communication Quarterly, 92(1), 121–141. https://doi. org/10.1177/1077699014558554 Brantner, C., Lobinger, K., & Wetzstein, I. (2011). Effects of visual framing on emotional responses and evaluations of news stories about the Gaza conflict 2009. Journalism & Mass Communication Quarterly, 88(3), 523–540. https://doi.org/10.1177/107769901108800304 Brecht, H., Deichmann, U., & Wang, H. G. (2013). A global urban risk index. The World Bank. https://doi.org/10.1596/1813-9450-6506 Brubaker, P. J., & Wilson, C. (2018). Let’s give them something to talk about: Global brands’ use of visual content to drive engagement and build relationships. Public Relations Review, 44(3), 342–352. https://doi.org/10.1016/j.pubrev.2018.04.010 Buntain, C. L., & Lim, J. R. (2018). #pray4victims: Consistencies in response to disaster on Twitter. Proceedings of the ACM on Human–Computer Interaction, 2, 1–18. https://doi. org/10.1145/3274294 CARE International. (2018). Suffering in silence: The 10 most under-reported humanitarian crises of 2017. https://reliefweb.int/sites/reliefweb.int/files/resources/72486f6-7858-02072019_ Report_Suffering-I.pdf Centre for Research on the Epidemiology of Disasters. (2020). Emergency events database (EM-DAT): The international disaster database. https://www.emdat.be/ Chouliaraki, L. (2006). The spectatorship of suffering. SAGE. Chung, S., & Lee, S. Y. (2019). Visual CSR messages and the effects of emotional valence and arousal on perceived CSR motives, attitude, and behavioral intentions. Communication Research, 46(7), 926–947. https://doi.org/10.1177/0093650216689161 Coombs, W. T., & Holladay, S. J. (2011). An exploration of the effects of victim visuals on perceptions and reactions to crisis events. Public Relations Review, 37(2), 115–120. https:// doi.org/10.1016/j.pubrev.2011.01.006 Dhanesh, G. S., & Rahman, N. (2021). Visual communication and public relations: Visual frame building strategies in war and conflict stories. Public Relations Review, 47(1), Article 102003. https://doi.org/10.1016/j.pubrev.2020.102003 Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4), 51–58. https://doi.org/10.1111/j.1460-2466.1993.tb01304.x Erjansola, A.-M., Lipponen, J., Vehkalahti, K., Aula, H.-M., & Pirttilä-Backman, A.-M. (2021). From the brand logo to brand associations and the corporate identity: Visual and identity- based logo associations in a university merger. Journal of Brand Management, 28(3), 241– 253. https://doi.org/10.1057/s41262-020-00223-5 Fahmy, S., Kelly, J. D., & Kim, Y. S. (2007). What Katrina revealed: A visual analysis of the hur- ricane coverage by news wires and U.S. newspapers. Journalism & Mass Communication Quarterly, 84(3), 546–561. https://doi.org/10.1177/107769900708400309 Fisher, D., Hagon, K., Lattimer, C., O’Callaghan, S., Swithern, S., & Walmsley, L. (2018). World disasters report: Leaving no one behind. International Federation of Red Cross and Red Crescent Societies. https://reliefweb.int/sites/reliefweb.int/files/resources/B-WDR- 2018-EN-LR-compressed.pdf Foroudi, P. (2019). Influence of brand signature, brand awareness, brand attitude, brand reputation on hotel industry’s brand performance. International Journal of Hospitality Management, 76, 271–285. https://doi.org/10.1016/j.ijhm.2018.05.016 22 Journalism & Mass Communication Quarterly 00(0) Gale, A., & Lewis, M. B. (2019). When the camera does lie: Selfies are dishonest indicators of dominance. Psychology of Popular Media Culture, 94, 447–455. https://doi.org/10.1037/ ppm0000260 Geise, S. (2017). Visual framing. In P. Rössler (Ed.), The international encyclopedia of media effects (pp. 1–12). John Wiley. Government of India. (2021). Lines of credit for development projects. https://mea.gov.in/ Lines-of-Credit-for-Development-Projects.htm#:~:text=Development%20assistance%20 in%20the%20form, been%20extended%20to%2064%20countries Guidry, J. P., Jin, Y., Orr, C. A., Messner, M., & Meganck, S. (2017). Ebola on Instagram and Twitter: How health organizations address the health crisis in their social media engagement. Public Relations Review, 43(3), 477–486. https://doi.org/10.1016/j.pubrev.2017.04.009 Hardin, M., Lynn, S., Walsdorf, K., & Hardin, B. (2002). The framing of sexual difference in SI for Kids editorial photos. Mass Communication & Society, 5(3), 341–359. http://doi. org/10.1207/S15327825MCS0503_6 Hellmueller, L., & Zhang, X. (2019). Shifting toward a humanized perspective? Visual fram- ing analysis of the coverage of refugees on CNN and Spiegel Online before and after the iconic photo publication of Alan Kurdi. Visual Communication. Advance online publica- tion. https://doi.org/10.1177/1470357219832790 Hong, H. (2013). The effects of human interest framing in television news coverage of medical advances. Health Communication, 28(5), 452–460. https://doi.org/10.1080/10410236.201 2.693013 Houston, J. B., Pfefferbaum, B., & Rosenholtz, C. E. (2012). Disaster news: Framing and frame changing in coverage of major U.S. natural disasters, 2000–2010. Journalism & Mass Communication Quarterly, 89(4), 606–623. https://doi.org/10.1177/1077699012456022 Javornik, A., & Mandelli, A. (2012). Behavioral perspectives of customer engagement: An exploratory study of customer engagement with three Swiss FMCG brands. Journal of Database Marketing & Customer Strategy Management, 19(4), 300–310. http://doi. org/10.1057/dbm.2012.29 Jennings, R. (2017, December 22). China is giving more foreign aid than it gets. Forbes. https:// www.forbes.com/sites/ralphjennings/2017/12/22/china-is-giving-more-foreign-aid-than- it-gets/?sh=256622af4f35 Jiang, H., Luo, Y., & Kulemeka, O. (2016). Social media engagement as an evaluation barom- eter: Insights from communication executives. Public Relations Review, 42(4), 679–691. https://doi.org/10.1016/j.pubrev.2015.12.004 Joye, S. (2009). The hierarchy of global suffering. Journal of International Communication, 15(2), 45–61. http://doi.org/10.1080/13216597.2009.9674750 Joye, S. (2010). News media and the (de)construction of risk: How Flemish newspapers select and cover international disaster. Catalan Journal of Communication & Cultural Studies, 2(2), 253–266. https://doi.org/10.1386/cjcs.2.2.253_1 Kim, C., & Yang, S. U. (2017). Like, comment, and share on Facebook: How each behavior differs from the other. Public Relations Review, 43(2), 441–449. https://doi.org/10.1016/j. pubrev.2017.02.006 Krippendorff, K. (2004). Content analysis: An introduction to its methodology (2nd ed.). SAGE. Ladrido-Ignacio, L., & Perlas, A. P. (1995). From victims to survivors: Psychosocial interven- tion in disaster management in the Philippines. International Journal of Mental Health, 24(4), 3–51. https://doi.org/10.1080/00207411.1995.11449321 Lee et al. 23 Lang, A. (2006). Using the limited capacity model of motivated mediated message processing to design effective cancer communication messages. Journal of Communication, 56(Suppl. 1), S57–S80. https://doi.org/10.1111/j.1460-2466.2006.00283.x Lang, S. (2013). NGOs, civil society, and the public sphere. Cambridge University Press. Li, Y., & Xie, Y. (2020). Is a picture worth a thousand words? An empirical study of image content and social media engagement. Journal of Marketing Research, 57(1), 1–19. https:// doi.org/10.1177/0022243719881113 Lim, J. R., Liu, B. F., Egnoto, M., & Roberts, H. A. (2019). Individuals’ religiosity and emo- tional coping in response to disasters. Journal of Contingencies and Crisis Management, 27(4), 331–345. https://doi.org/10.1111/1468-5973.12263 Miller, A., & LaPoe, V. (2016). Visual agenda-setting, emotion, and the BP oil disaster. Visual Communication Quarterly, 23(1), 53–63. https://doi.org/10.1080/15551393.2015.1128335 Neuendorf, K. A. (2010). Reliability for content analysis. In A. B. Jordan, D. Kunkel, J. Manganello, & M. Fishbein (Eds.), Media messages and public health: A decision approach to content analysis (1st ed., pp. 67–85). Routledge. Olteanu, A., Vieweg, S., & Castillo, C. (2015). What to expect when the unexpected happens: Social media communications across crises. In Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing (pp. 994–1009). Association for Computing Machinery. https://doi.org/10.1145/2675133.2675242 Ordenes, F. V., Grewal, D., Ludwig, S., Ruyter, K. D., Mahr, D., & Wetzels, M. (2019). Cutting through content clutter: How speech and image acts drive consumer sharing of social media brand messages. Journal of Consumer Research, 45(5), 988–1012. https://doi.org/10.1093/ jcr/ucy032 Organisation for Economic Co-operation and Development. (2020). Aid by DAC members increases in 2019 with more aid to the poorest countries. https://www.oecd.org/dac/financ- ing-sustainable-development/development-finance-data/ODA-2019-detailed-summary.pdf Panofsky, E. (1970). Meaning in the visual arts. Penguin. Parry, K. (2011). Images of liberation? Visual framing, humanitarianism and British press photography during the 2003 Iraq invasion. Media, Culture & Society, 33(8), 1185–1201. https://doi.org/10.1177/0163443711418274 Powell, T. E., Boomgaarden, H. G., De Swert, K., & de Vreese, C. H. (2015). A clearer picture: The contribution of visuals and text to framing effects. Journal of Communication, 65(6), 997–1017. https://doi.org/10.1111/jcom.12184 Powers, M. (2018). NGOs as newsmakers: The changing landscape of international news. Columbia University Press. PRovoke. (2017). International organizations increasingly use visual content to engage fol- lowers. https://www.provokemedia.com/latest/article/international-organizations-increas- ingly-use-visual-content-to-engage-followers Rebich-Hespanha, S., Rice, R. E., Montello, D. R., Retzloff, S., Tien, S., & Hespanha, J. P. (2015). Image themes and frames in U.S. print news stories about climate change. Environmental Communication, 9(4), 491–519. https://doi.org/10.1080/17524032.2014.983534 Riffe, D., Lacy, S., Watson, B. R., & Fico, F. (2019). Analyzing media messages: Using quan- titative content analysis in research (4th ed.). Routledge. Rodriguez, L., & Dimitrova, D. V. (2011). The levels of visual framing. Journal of Visual Literacy, 301(1), 48–65. https://doi.org/10.1080/23796529.2011.11674684 Rogers, S., & Thorson, E. (2000). “Fixing” stereotypes in news photos: A synergistic approach with the Los Angeles Times. Visual Communication Quarterly, 7(3), 8–11. https://doi. org/10.1080/15551390009363436 24 Journalism & Mass Communication Quarterly 00(0) Saxton, G. D., & Waters, R. D. (2014). What do stakeholders like on Facebook? Examining public reactions to nonprofit organizations’ informational, promotional, and community- building messages. Journal of Public Relations Research, 26(3), 280–299. https://doi.org/ 10.1080/1062726X.2014.908721 Schlimbach, H. J. (2013). How disaster relief organizations solicit funds: The effects of disaster presence, message framing, and source credibility on an individual’s intention to donate [Master’s thesis, University of Texas at Austin]. UT Electronic Theses and Dissertations. Seo, H., & Vu, H. T. (2020). Transnational nonprofits’ social media use: A survey of com- munications professional and an analysis of organizational characteristics. Nonprofit and Voluntary Sector Quarterly, 49(4), 849–870. https://doi.org/10.1177/0899764020908340 Smith, B. G., & Gallicano, T. D. (2015). Terms of engagement: Analyzing public engage- ment with organizations through social media. Computers in Human Behavior, 53, 82–90. https://doi.org/10.1016/j.chb.2015.05.060 Spence, P. R., Lachlan, K. A., Lin, X., & del Greco, M. (2015). Variability in Twitter content across the stages of a natural disaster: Implications for crisis communication. Communication Quarterly, 63(2), 171–186. http://doi.org/10.1080/01463373.2015.1012219 Stoddard, A. (2003). Humanitarian NGOs: Challenges and trends. In J. Macrae & A. Harmer (Eds.), Humanitarian action and the “Global War on Terror”: A review of trends and issues, HPG Report 14 (pp. 25–36). Overseas Development Institute. https://cdn.odi.org/ media/documents/287.pdf Thrall, A. T., Stecula, D., & Sweet, D. (2014). May we have your attention please? Human- rights NGOs and the problem of global communication. International Journal of Press/ Politics, 19(2), 135–159. https://doi.org/10.1177/1940161213519132 Twitter. (2020). Get Tweet timelines. https://developer.twitter.com/en/docs/twitter-api/v1/ tweets/timelines/api-reference/get-statuses-user_timeline van Grinsven, G. B., & Das, E. (2016). Logo design in marketing communications: Brand logo complexity moderates exposure effects on brand recognition and brand attitude. Journal of Marketing Communications, 22(3), 256–270. https://doi.org/10.1080/13527266.2013.8 66593 van Leeuwen, T. (2001). Semiotics and iconography. In T. van Leeuwen & C. Jewitt (Eds.), Handbook of visual analysis (pp. 92–118). SAGE. Wallemacq, P., & House, R. (2018). Economic losses, poverty, & disasters, 1998–2017. Centre for Research on the Epidemiology of Disasters, United Nations Office for Disaster Risk Reduction. https://www.undrr.org/publication/economic-losses-poverty-disas- ters-1998-2017 Walter, E., & Gioglio, J. (2014). The power of visual storytelling. McGraw-Hill. World Confederation of Physical Therapy. (2019). Organisations involved in disaster manage- ment. https://web.archive.org/web/20180324132316/https://www.wcpt.org/disaster-man- agement/Organisations-involved-in-disaster-management Wozniak, A., Lück, J., & Wessler, H. (2015). Frames, stories, and images: The advantages of a multimodal approach in comparative media content research on climate change. Environmental Communication, 9(4), 469–490. https://doi.org/10.1080/17524032.2014.9 81559 Yang, Z. J., McComas, K. A., Gay, G., Leonard, J. P., Dannenberg, A. J., & Dillon, H. (2010). Motivation for health information seeking and processing about clinical trial enrollment. Health Communication, 25(5), 423–436. https://doi.org/10.1080/10410236.2010.483338 Lee et al. 25 Yang, Z. J., McComas, K. A., Gay, G., Leonard, J. P., Dannenberg, A. J., & Dillon, H. (2011). Information seeking related to clinical trial enrollment. Communication Research, 38(6), 856–882. https://doi.org/10.1177/0093650210380411 Zhang, X., & Hellmueller, L. (2017). Visual framing of the European refugee crisis in Der Spiegel and CNN International: Global journalism in news photographs. International Communication Gazette, 79(5), 483–510. https://doi.org/10.1177/1748048516688134 Zillmann, D., & Brosius, H.-B. (2000). Exemplification in communication: The influence of case reports on the perception of issues. Routledge. Author Biographies Sun Young Lee (PhD, University of North Carolina–Chapel Hill) is an assistant professor in the department of communication at the University of Maryland–College Park. Her research focuses on corporate social responsibility (CSR) and crisis/risk communication. In particular, she investigates visual strategies in various contexts, including CSR and international crises, how co-creational strategies for CSR activities can generate social value, how publics process CSR messages, and the effects of CSR strategies before or after a crisis. JungKyu Rhys Lim (PhD, University of Maryland–College Park) is a digital media behavioral scientist at the World Bank. He studies behavioral interventions to help individuals, communi- ties, and organizations better prevent, prepare for, respond to, and recover from risks and crises. His research utilizes mixed methods, including quantitative, qualitative, and computational methods. Duli Shi (MA, University of Maryland–College Park; M.S., South China University of Technology) is a Ph.D. candidate in the department of communication at the University of Maryland–College Park. Her research focuses on the effective communication of corporate social responsibility (CSR), as well as the strategic role of public relations in organizations and in the broader society.