In the Eye of the Storm: Social Media and Crisis Management

Abstract

Social media, also called Web 2.0, is a generic term used to talk about applications that allow users to create, manipulate, and disseminate content as much as possible in real time. These applications allow for several possibilities that range from involvement to participation, communication, and collaboration of users. They allow everyone with minimal access to the Internet to publish, share, review, comment, and post items, such as mentions, comments, information, videos, and photos. In a crisis, social media becomes a double-edged sword. It can play an essential role during the prodromal, acute, chronic, and resolution phases of natural disasters and human-made crises. Social media can also be at the origin of the crisis or the reason for its amplification. Social media facilitates an increase of interactions between main actors at the center of a crisis. This chapter combines social media content analysis (opinion detection and sentiment analysis) with network analysis (ego network analysis) and nodes centrality assessment to critically evaluate how social media affects the crisis management process.

1. Introduction

The Internet has taken only a quarter of the time it took television channels to become popular. Users themselves set the trends and shape the framework in which organizations interact with them. Networks are driving the evolution of ever shorter messages, delivered with increasing frequency through a plurality of means and channels. Social media, also called Web 2.0, is a generic term used to refer to applications that allow their users to create, manipulate, and distribute content as much as possible in real time. These applications allow for several possibilities that range from involvement to participation, communication, and collaboration of users. They allow everyone with minimal access to the Internet to publish, share, review, comment, and post items, such as mentions, comments, information, videos, and photos. In this way, social media has distinguished itself from traditional websites. Kaplan et al. [1] have defined social media as “a group of internet-based applications that build on the ideological and technological foundations of Web 2.0 and allow the creation and exchange of user-generated content.”

Indeed, with social media, the user can generate their own content, unlike with websites where they can only view it [2]. Social media companies have unleashed persuasive platforms that generate useful data for public debates, socialization, and information exchange, increasing their impact on daily lives [3]. Therefore, social media has substantially increased communication channels by allowing a constant interaction of information and opinions on a one-to-one, one-to-many, and many-to-many communicative basis [1, 4]. Social media has made it possible to blur boundaries and time by articulating ubiquity and multisharing. A story can spread virally to millions of people in just a few minutes [4]. Social media is characterized by interactivity, user-generated content, and multidirectional communication flows. With social media, the immediacy of information takes a new form as it arrives in real time. The public is no longer made of the distant spectators they once were. They do their own research on the authenticity of the news and sometimes influence the course of an event. In this sense, social media blurred the lines between the public and the audience. While for a long time, traditional media users acted as an audience (i.e., a passive spectator of the content diffused from these media), social media users now act as a public (i.e., they can influence the course of events through their interactions and have agency over events). Almost 5 billion people are connected to the Internet, and more than half of humanity (4.2 billion people) was connected to social media in 2021. The number of social media users grows by almost 13% yearly [5]. The progress of social media is the result of globalization. Ulrich Beck, a German sociologist, links globalization and crisis. In his famous theory on risk society, Beck argued that risk is inherent to modern society. This concept highlights social media’s role as a bridge between globalization and crises. Social media has influenced crisis management in the health, corporate, nonprofit, religious, political, and natural disaster sectors [6].

As far as crises are concerned, just in 2021, a total of 432 catastrophic events were recorded (Figure 1). Floods dominated these events, with 223 occurrences, up from an average of 163 annual floods recorded across the 2001–2020 period. Storms were the second most frequently recorded disaster, with 121 events recorded in 2021¹.

Figure 1.

This figure demonstrates the top 10 economic losses and disaster trends (source: 2021 disasters in numbers).

Crisis management phases	Number of articles
Mitigation	11% (n = 24)
Preparedness	23% (n = 50)
Response	57% (n = 125)
Recovery	9% (n = 20)

Table 1.

Breakdown of the corpus based on crisis management phases.

These numbers do not account for technical and human-made crises; if we were to add them, the tally will be higher. In fact, the more people use social media, the more researchers and practitioners need to understand social media’s impacts during a crisis [7]. This chapter focuses on the links between social media and crises, through the following research question: How is social media used in crisis management? Before responding to this question, the chapter will first outline the applied methodological approach, then explain critical features of social media that are involved in crisis management, and finally describe the two primary analysis trends of crises on social media that are regularly presented in the literature: the content analysis and the network analysis.

2. Use of social media in crisis management

When a crisis strikes, the team in charge of responding is always in need of tools to manage the overflow of information by summarizing it. Time being of the essence, it is critical to collect, review and classify high-value messages while using them to identify and manage issues as they arise or are amplified [8]. Social media has developed features that allow the crisis management team to perform disaster forensics [9].

Social media is essential throughout the crisis management process, from early detection to recovery. Social media has increased the volume of connections and the proactive nature of exchanges, which actively participate in a new form of crisis management [10]. This is mainly due to the trust relationship that social media has created with traditional media [11]. Thus, according to Timothy Coombs [12], researchers in crisis communication spotted the potential of social media long before traditional communications research [13]. Social media allows for an increase in the interactions of key actors at the center of a crisis, such as citizens, local communities, and officials from different levels of government [14]. This configuration is found in a crisis because people often are confined to hiding places, constrained by, among other things, the triggering of barricaded confinement. Many used social media to communicate with their relatives and reassure them of their conditions. However, what interests us here is the social media activity that manages this crisis. Thus, as Rainer et al. [15] have demonstrated, social media can be used to allow for effective crisis management in several aspects. First, the management of risk and the prediction of its occurrence. Indeed, monitoring social media in real time increases the speed and efficiency with which emergency managers can react. Qualitative and quantitative processing of social media content increases the likelihood of predicting mass behavior and better understanding potential risks [16].

Social media also allows for better risk analysis. The information gathered helps first responders assess the danger of the situation, from the sharing of information and knowledge about the site on social media to the extent of the damage and the number of victims. This allows response teams to adequately plan for specific resources and determine the appropriate scope of work before starting tasks that need to be done, even before these teams arrive on the scene.

Once a crisis begins, social media can be used to coordinate the efforts of different working groups. The crisis management team members can be tasked according to their skills and the needs in the field. Social media has become an essential tool in the organization of post-crisis aid and allows the coordination of different community organizations that want to help with relief efforts. By providing adequate and real-time information to people affected by the crisis, social media enable people to understand the full scope of the crisis and signal the presence of crisis managers.

The rapid and real-time dissemination of information is undoubtedly one of the most used functions of social media in times of crisis. It can help save lives. In addition, it offers the advantage of reaching people who are cut off from traditional means of communication, especially in developing countries, where there are often more mobile phones than landlines and more smartphones and handhelds than desktop computers.

Social media also puts the role of first witnesses back at the center of the intervention. These act as citizen journalists and transmit valuable information with regular updates on the situation on the ground. Responders cannot be everywhere and often have limited resources [17].

Web 2.0 reorients the narrative about the crisis and the public at the heart of it, defining the causes, identifying the issues, and predicting the consequential impact [18]. Austin et al. [19] reflected on an organization’s interactions over social media with its various audiences that produce, consume, and disseminate information during a crisis. Their model, called crisis communication through social media, has identified three types of social media to consider before, during, and after a crisis. Some influential creators create content and convey information about the situation that others consume. Followers consume the information provided by the creators. Inactive participants eventually use the data from the creators through word-of-mouth or through traditional media, which also follows creators. Researchers have recognized Twitter as playing a significant role during natural disasters.

Social media like Twitter provides a platform for the efficient organization of relief efforts and for strategic planning to avoid further human and material damage [20, 21, 22]. Web 2.0 technologies, social media, and media data mining are new technological forms that provide and gather information about the population affected by a crisis in an efficient manner [23, 24]. Social media help increase the flow of information between people directly affected and those close to them (family and friends). Most platforms have features that allow a user to let their loved ones know they are safe when a disaster occurs. This online reporting reassures others and better channels the search and rescue of those who need it most. Social media in crisis contexts allow for rapid updates on developments and continuous interaction with the community and neighborhood [25, 26].

In a crisis involving a sniper on the loose, it is wiser to use a phone to connect to social media than to speak in person, which would risk alerting the sniper of someone’s presence and attempting to call for safety. This was the situation in the shootings of May 14, 2022, in Buffalo, New York, and, in the terrorist attack of September 21, 2013, at the Westgate shopping mall in Nairobi, Kenya. In these situations, it appears as if texting and updating the status through social media provided information to relatives while it is safer to do so than calling and using the voice, which could result in letting the sniper where you are hiding. In the case of the attack of January 9, 2015, against the kosher grocery store in the suburbs of Paris, France, social media have helped in crowdsourcing with people on the ground and played a key role in supplementing the 24 h/7 television cycle to contribute to the increase of the post-stress resulting from the trauma of being exposed physically to those events [27, 28].

Now that types of social media and users and their practicality during a crisis have been identified, and the next section will examine how social media can be used during a crisis. The answer to this how follows the phases that evolve chronologically during a crisis.

3. Methodology framework

This research uses the convergent parallel design method to triangulate multiple data sources using quantitative and qualitative approaches (QUAN +QUAL). From this, links between two phenomena such as crises and social media can be analyzed, and further interpret the combined results [29]. This approach comes from one of the most regularly used studies concerning social media [30]. This chapter conducts a systematic review of literature about crises and social media to determine the trends. A literature review, which examines existing research and information on a chosen field of study, is a crucial component of a research study. Fink’s [31] definition of a research literature review as “a systematic, explicit, and reproducible method for identifying, evaluating, and synthesizing the existing body of completed and recorded work produced by researchers, scholars, and practitioners” corresponds to what we understand as a systematic literature review (p. 3). A systematic review is also a type of literature review. The main difference between a literature review and a systematic review is their focus on the research question; a systematic review is focused on a specific research question, whereas a literature review is not. The main objective of this type of research is to identify, review, and summarize the best available research on a specific research question. Systematic reviews are primarily used because reviewing existing studies is often more practical than conducting new research [32]. As adopted in this chapter, the review is predominantly descriptive to synthesize the critical aspects of using social media in crisis management.

The methodology followed was inspired by the work of Snelson [30], notably the four-stage approach, as well as the work of Kankanamge et al. [33], to highlight a comprehensive understanding of the use of social media in crisis (natural and man-made) management. The scientific articles used as a corpus in this chapter were selected from databases using four stages: pre-search, data collection, data cleaning, and analysis.

3.1 Data identification and collection

The research for this chapter was conducted from April to May 2022. Using Boolean search with keywords, such as social media and crisis management, articles were selected from the following databases: Academic Search Premier (n = 458); Applied Social Sciences Index and Statistics (n = 989); ProQuest (n = 370); Science Direct (n = 451); and Scopus (n = 693). These databases were chosen for their broad reach across multidisciplinary and interdisciplinary peer-reviewed journals. They also represent key research fields currently popular in scholarly publications. In studying links between social media and crisis, it is critical to research both the text and the context [34].

New specific keywords were included, such as social media and natural disasters, social media and organizational crisis, social media and reputation management, social media and image repair, crisis informatics, social media crisis communication, social media, and financial crisis, social media and political crisis, social media and economic crisis, social media and social crisis, and social media and scandal to reduce the scope of the research. The search was also restricted to the classic stages of crisis management as identified by Bundy et al. [35] and Lai & Wong, [36]: pre-crisis, crisis, and post-crisis. The articles retained for analysis were the ones with Boolean phrases, such as social media and crisis mitigation, social media and crisis preparedness, social media and crisis response, and social media and crisis recovery (Table 1). This research was also conducted to understand the influence of social media on the five “critical tasks” for leaders during crises, as identified by Boin et al. [37], namely: sense-making, decision-making, meaning-making, terminating, and learning.

3.2 Data cleaning

After filtering social media streams and reducing irrelevant information, the reference was identified by checking the title, abstract, key findings, and conclusion to ensure that each selected article was related to the research question. This methodological discrimination process brought down the total tally of articles from all databases to 219. Here is the following classification of articles based on the crisis management process:

3.3 Results

Content analysis was used to assess the corpus. “Content analysis is a research technique that is based on measuring the amount of something (violence, negative portrayals of someone, etc.) in a representative sampling of some mass-mediated popular form of art” ([38], p. 25). The “content” refers to words, meanings, pictures, symbols, ideas, themes, or any message that can be communicated. In content analysis, a systematic sample of texts is used in the study, and classification systems are devised to identify different features of the text, which are then counted with an emphasis on objectivity and reliability. This is done to describe the substance characteristics of message content, to describe the form characteristics of message content, to make inferences about audiences of content, and to predict the effects of content on audiences [39].

In the case of this research, content analysis was mainly used to find meaningful links between social media and the crisis management process, which prompted an assessment of the reliability of those two factors in the corpus using the kappa Cohen coefficient [40]. Cohen’s kappa coefficient is a statistical measurement that calculates the level of agreement between two variables according to a formula. As a general rule, kappa is always less than or equal to 1 [41]. When kappa is less than or equal to 0.20, the level of agreement between the variables is poor. This agreement is fair when the kappa is between 0.20 and 0.40; it is moderate if kappa is between 0.40 and 0.60; it is good when the kappa is between 0.60 and 0.80, and it is very good if the coefficient is between 0.80 and 1. The coding of the corpus was conducted using NVivo and R software until an 0.80 coefficient was obtained. Case studies are by far the most used approach in the corpus, with 64% (n = 140), while for data collection techniques, the content analysis, including literature reviews, accounts for 47% (n = 103). Next comes focus groups (28% n = 61), interviews (12% n = 26), surveys (8% n = 17), and finally fieldwork and observations (5% n = 11).

The corpus studied shows Twitter as the most used platform representing 43%, highlighting its real-time capacities to provide useful data as the crisis outspread, thus improving situation awareness [42]. Some applications, such as WhatsApp, are also widely used and are part of the 15% of the corpus represented under the terms specialized platforms. In most cases, the literature revealed the use of combined platforms, such as Twitter and Facebook, Facebook and Instagram, or links from the three abovementioned platforms to YouTube. It is worth mentioning that the total of platforms used surpassed 100% because some studies in the selected corpus combined platforms for their research.

As per the platforms, the results were observed as shown in Figure 2.

Figure 2.

List of social media platforms represented in the corpus.

Natural disaster	Country	Year	Application	Authors
Hokkaido Earthquake	Japan	2018	K-DiPS	Nakai et al. [112]
Lombok Earthquake	Indonesia	2018	Crowhelp	Rachmah et al. [113]
Gorkha Earthquake	Nepal	2015	Urep	Goda et al. [114]
Haiti Earthquake	Haiti	2010	Ushahidi	Yates & Paquette [115]
Chile Earthquake and Tsunami	Chile	2010	Ushahidi
Wenchuan Earthquake	China	2008	WebGIS	Huang et al. [116]
Saguenay Floods	Canada	1996	Hazus (web application, not mobile)	Natev & Todorov [107]

Table 2.

Non-exhaustive list of web applications playing a key role during natural disasters as per the corpus.

Based on these findings, the literature provides tools and processes to assess data during the crisis on Facebook and Twitter, the most used platforms.

3.3.1 Data analysis through twitter

Twitter is a tool that has not always been used to its full potential during crises. Indeed, if citizens are massively present, their number increases substantially during a crisis, and some governments have often been reluctant to use this tool [43]. In addition to all the data identified above for Facebook and Twitter, during a crisis, it is the place from which rumors start that can impact the organization’s ability to resolve the crisis and, later, its reputation. This is why the analysis of network interaction seems the most interesting. To determine this, several researchers use NodeXL, a network in the form of a list point to symbolize the relationship that exists not only between Twitter users but especially their interaction on a particular piece of content [44, 45, 46]. Each action performed on Twitter leaves traces that form a network. NodleXL makes it possible to highlight the connection networks of Twitter users. This tool uses the Clauset-Newman-Moore algorithm to divide network users into subgroups and generate graphs that identify connections between users [45].

Twitter’s communication interface is quite flexible and allows for the exchange and communication through words with a more significant number of users, even if they do not follow each other. They are following each other and are not in the same network. Accounts on Twitter are both public, that is, visible to all, including those who are not registered as members, or else private, that is, visible only to those in one’s network [47, 48]. During a crisis, information is gathered by tracking keywords and tweets made from those words. This was seen during the Boston terrorist attack with the keyword #Bostonstrong, among others.

Twitter allows access to public tweets through two practical tools: the application programming interface (API) and streaming, enabling one to see the tweets at any time and see them as they are written. There can be three types of analysis: activity analysis, network analysis, and content analysis [47].

The calculation of statistics and indicators describing the activities Twitter activity captured in a dataset at a particular time relies primarily on processing these datasets based on specific communication profiles. The additional filtering of the data during a certain time, user, or keyword evaluation may also be necessary to better understand the activity on Twitter better.

The analysis of Twitter activity can be done temporarily. It then includes the overall volume of tweets plus the hourly volume of different types of tweets over time (original tweets, replies, unpublished retweets, revised retweets, tweets containing URLs, etc.). This analysis is also based on the volume of specific keywords (or clusters of keywords over time), the number of active users over some time (day, hour, day of the week) period (day, hour, minute), and the average number of tweets per user over a while.

3.3.2 Data analysis through Facebook

There are four primary actions on Facebook. You can either post a piece of information or a video, like someone’s post, comment on it or share it with your network of friends and admirers. The ability to spread is one of the things to measure with Facebook accounts. Outreach is an aggregation of the total number of pages, likes, friends, and people talking about it, those who spread the information back in turn, and the click-through rate obtained [49]. From a post, one can evaluate the viral nature of the data by determining the number of people who have shared it. This allows us to know the type of post that pushes the audience to action, however small it may be. Obtaining such data is crucial in post-crisis analysis because it gives an idea of the quality of information to look for in times of crisis [50, 51]. Applications such as Facebook Insights also provide information about page views by unique visitors, that is, those who came directly to the Facebook page without being driven there by an ad.

The engagement rate is generally one of the most calculated data on Facebook. It measures how well an organization interacts with its audience on Facebook. It is data that is calculated post-ante. It is an equation that divides the number of people who clicked, liked, commented, or shared the post by the number of people who have seen it. This aspect is critical during a crisis when you want to calculate the relevance of the diffusion of information [52]. It is necessary to transcend the reach rate (the number of people who have seen the post) because many accounts are fake or robotic and automatically generate traffic without the expected impact. The rate is a useful measure to evaluate the quality of the information content generated on Facebook. Only the account administrator can obtain it.

4. Key features of social media in crisis management

Several essential aspects of social media need to be assessed to understand its role in managing crises properly. One of these crucial aspects is the concept of a network, which, when studied, helps indicate how information is created.

4.1 Networks

Throughout history, networks have always been part of society, existing at the center of power [53]. They suggest how people have lived in society through their form and how people are connected through various positions. Indeed, people seem tailored to influence and interact with one another, and our brains seem wired to seek connectivity with others [54]. Linking globalization and the evolution of the information and communication technology of (ICT), Manuel Castells [55] explained that new relations using ICT have emerged, creating a new social structure. The analysis of social ties is a field that has grown in importance while establishing the sociology of information and communication techniques [56]. A social network is a social structure composed of individuals (or organizations) called “nodes,” linked by one or more specific types of interdependence, such as friendship; kinship; common interest; financial exchange; aversion; sexuality; or relationships of belief, knowledge, or prestige [57]. Scholarly literature has shown that it is impossible to manage a crisis through social media without a social network analysis (SNA), which refers to the regularities in the patterning of relationships among individuals, groups, or organizations. When a crisis arises, it is critical to understand how information circulates within a network, wherever the information has started or if social media simply amplify it.

4.2 Homophily

Although definitions of homophily differ in their specificity, their basic theoretical intent agrees that it is “the degree to which interacting pairs of individuals are similar in certain attributes” ([58], p. 402). Homophily is also “the tendency of individuals to interact with similar individuals” ([59], p. 463). Fincham explained that homophily acquires meaning for the individuals involved and thus influences their social interaction (2019). One of the important observations made by social scientists is this tendency in social groups to connect similar people (after all, birds of a feather flock together). This tendency significantly affects the values derived from social media, where people often encounter similar voices and interact with like-minded people. Homophily has predictive power in social media, so researchers can predict real-life friendships by looking at online interactions, common interests, and location [60]. This concept can be seen as correlated with the question: Who social media users view to be trustworthy in a crisis? It explains how information circulates during a crisis and key phenomena such as fake news and echo chambers.

4.3 Bubble filters

Bubble filter is the concept by which algorithms track your activities online and systematically arrange the content you have access to base on your preferences. These filters structure people’s online experience, isolating them from the information they have not previously expressed interest in [61]. The bubble filters are based on the selective exposure theory [62]. When managing a crisis, bubble filters represent a serious threat in the CMT attempt to reach out to targeted audiences through social media as the filters might not allow the information to go through their bubble of preferences [63]. While networks, homophily, and bubble filters are hindering the reach, echo chambers’ next concept impacts the message’s persuasiveness during a crisis.

4.4 Echo chambers

Integral to the echo chamber is the idea that opinions in the network are polarized with self-reinforcing nodes [64]. Online users decide to isolate themselves in the worlds of a shared imagination [65]. Recent studies have shown that online users tend to select the information that adheres to their belief systems, ignore information that does not, and join groups—that is, echo chambers—centered around a shared narrative. This concept can explain how it was difficult during the COVID-19 pandemic to present rational arguments against conspiracy theories [66, 67, 68, 69]. Information circulation is not the only critical aspect of social media related to a crisis; the speed of information travel is also essential. Other key features that link social media to a crisis include virality and buzz.

4.5 Virality

The concept of virality is borrowed from medicine and related to the proliferation of harmful viruses inside the human body, or in this case, society. Metcalfe’s law [70], which famously characterizes many of the effects seen in communication technologies and networks, from the Internet and social networking to the World Wide Web, could explain virality as a dynamic of information transmission speed on social media. The law states that the value of a telecommunications network is proportional to the square of the number of connected users of the system. According to Huberman et al. [71], virality is related to mimetics since it boils down to sharing and looking for information that other people are already looking for and sharing. People online are hungry for novelty and anything that sounds new.

4.6 Buzz

Crises on social media are by-products of bad buzz. This is another expression to designate electronic word-of-mouth that has the power to hinder an organization’s reputation. Since communication on social media happens in real time, most users have the privilege of being news breakers organizations managing crises should pay careful attention to that phenomenon.

Now that features of social media with the power to impact crises have been outlined, it is time to look at how social media, according to the literature, manage crises.

5. Crisis watch: social media monitoring

There are several factors to consider when monitoring the possibility of a crisis on social media. It involves tracking specific elements on the social media horizon to identify those that could be problematic. This monitoring provides crucial information in real time and on a continuous basis, covering controversial topics. At the same time, it gives access to new or hidden elements that emerge in conversations that an organization previously ignored, giving a whole new perspective on the subject.

Social media monitoring is an integral element of risk communication and, therefore, an essential component of any crisis communication strategy. Web 2.0 is where reputational risks are high, so it is advisable for 24/7 monitoring [72]. The purpose of monitoring is to identify radical opinions and elements that could damage an organization’s reputation and thus trigger a crisis [73] and identify all negative opinions in a general way [74]. In this way, social media monitoring can track the trajectory of comments and discussion patterns that emerge from different topics accompanying main comments, especially the profiles of the most influential people. Such a predisposition already draws the cartography of the action in a crisis. Time is the scarcest commodity during a crisis. The continuous monitoring of the networks allows one to know how to orient oneself during a crisis.

The social media monitoring process needs to be segmented into several steps to be effective [47, 75, 76]. First, the preparation is done based on objectives established beforehand in close connection with the organization’s mission. The proposal is not to monitor citizens’ private conversations but rather the information they have chosen to make public on social media. Although the first type of surveillance is espionage, the second type belongs to the field of strategic communication.

Preparing for a crisis takes an organization through the process of defining a crisis. Potential risks are identified for each organization in the crisis communication plan. The preparation phase also requires assessing the resources necessary to ensure adequate monitoring. Once preparation is done, the monitoring itself begins. At this stage, the main focus is on collecting critical data. As with any information-gathering exercise, it is essential to use an appropriate method. This can be done manually using Boolean operators or search engines [77]. The next step is to analyze the information. This can be done in a statistical, metric, or specific way on the activities, content, or network. This process is operationalized from the applications of the program interface designated API, which is a system of tools and resources in an operating system that allows an operating system to enable developers to create software applications such as Tweet Archivist, Tweetdeck, the Hootsuite galaxy of Hootsuite, Netvibes, and Trackur APIs [76]. Facebook and Twitter both offer APIs for data search, as do some free search engines such as Social Mention. Monitoring is an essential tool in the decision-making process in times of crisis [78, 79, 80]. In crisis cases, such as mass shootings, perpetrators have been vocal on social media for some time and could have been spotted with proper monitoring [81].

To summarize, following [76], a textual analysis that allows for understanding user opinions is necessary for monitoring. This requires specialists who work 24/7 by combining a manual approach with the appropriate software. Monitoring social media with essential information monitoring tools is critical in crisis management.

6. Response and recovery phases: social media analysis

The following section adopts a normative approach because, in the literature, the authors advised the crisis management team (CMT) on harvesting social media’s power to mitigate the crisis effects.

Social media increases information and knowledge during a crisis, whether the public is affected or not [82]. After social media monitoring comes social media analysis, which is the development and evaluation of tools to visualize, collect, and track social media data based on the particular requirements of a target application [83]. There are different types of data: The so-called structured data include the sociodemographic elements of the user; thematic data include content such as likes, comments, shares, and more; unstructured data can be stored and include, among other things, the identity of the ad, timestamp, username (of the author), the content of the posting, sometimes content, and occasionally even the type of posting. It is, therefore, advisable to adopt some practical approaches to social media analysis, such as thematic or trend-based analysis, opinion, and sentiment analysis, the structural method [84], or the ecosystem approach [85]. When responding to a crisis, collecting meaningful information and tracking its propagation are critical for a crisis management team [86].

Social media analysis during a crisis shows a clear difference in focus in the literature based on the research approaches where social science authors produce more qualitative research focusing on content analysis: opinion detection and sentiment analysis. In contrast, authors from the computers and health sciences fields emphasize quantitative research using structural methods with network analysis and computational methods as critical aspects of understanding how social media could be used to assess the crisis dynamic.

One of the key limitations in the literature is that there is little to no exchange and discussion between researchers and, more important, no comprehensive approach to analyses of response and recovery through social media. The following section will be divided into two main parts: It will first present the quantitative and then qualitative methods as they appear in the literature review.

6.1 Qualitative data analyses (QDA)

This section is about the research that displayed analysis of the impacts of social media on crises that mostly focus on text-based data. They describe the use of words, assess non-numeric data information, and explain the nature and deployment of key concepts and ideas related to the topic.

6.1.1 Content analysis

Opinion detection: During a crisis, social media can be used to gather necessary information, especially with the detection of opinions. The crisis management team struggled to convince people to abandon their homes in natural disasters such as the California wildfires in the United States. An opinion survey can help identify the fears and needs of people caught in the middle of a crisis like this one. Opinion research uses a thematic approach and requires an analysis process of social media content.

For this purpose, there are two main approaches [72]:

1. Natural language processing analyzes representations and implicit meanings based on a vector of texts and meanings, leading to identifying degrees of positive or negative opinions of texts produced in social media.

2. The semantic web approach detects explicit representations of the domain based on the semantic annotations that trace the ontology of the text from keywords or tags.

Both approaches focus on detecting opinions on social media to provide senior executives with a rational basis for deciding which topics are the determinants of important topics to monitor on social media. Opinion detection on social media has often been used to predict major trends in practice and acceptable organizational standards.

Some authors have found strong links between social media opinions and real-life events [8, 87, 88]. The detection of opinions on social media can be done through several steps: detection, tracking, classification, and verification—especially in a crisis involving rumors, such as bombing alerts and cyber crises [89].

First, one must identify the contributor (i.e., the one who is expressing the opinion). Second, the sentiments they express must be classified. According to the classic positive/negative/neutral approach, this classification highlights two essential aspects: polarization and information. The next step is the representation scale, a numerical determination of an opinion’s degree of negativity (−1) or positivity (+1). Finally, it is important to look at the purpose of the opinion: what subjects it deals with, who it is about, and what events it is about [90]. For each text, an opinion can be represented by rhetorical aspects, concepts used, and keywords [72]. These opinions are extracted as algorithms that combine the time at which the opinions were expressed, the average of the opinions on a particular topic, the average subjectivity of each opinion, the standard of positive and negative opinions, and the total number of views [90].

6.1.2 Sentiment analysis

Once opinion detection has been achieved, and the classification is done, managing a crisis through social media should lead to sentiment analysis, which is critical when managing situations such as recovery from natural disasters [91].

Sentiment analysis in social media, especially during a crisis, combines the factual elements in the text with the expression of emotions for an effective evaluation of content produced by both individuals and online communities [92, 93, 94]. This analysis combines phrases and keywords [90]. It performs a morphological analysis by determining the nouns and verbs used in each sentence using a software program, according to the following declination: a taxonomy of emotion-related words paired according to their common root. This allows a lexical determination based on verbal forms, for example. It is then essential to contextualize the use of the words [95]. These words may be related to feelings, but what they are meant to express may be different depending on the context and the collective understanding community in which they are used [96]. The root match here is no longer enough. Sentiment detection identifies comments found in texts and blogs and annotates them independently of the main articles to aggregate the sentiments appropriately. Sentiment grammar identifies these and then associates them with relevant targets and the owners of those opinions. Sentiment aggregation combines the scores of each feeling expressed on the networks [97].

To better analyze feelings on social media in a crisis, it is necessary to understand them through the dichotomy of normative and informative elements that govern the formation of an opinion. In the processing of a statement, the normative elements are the socialization of the person, their rank within the group and the group itself, and the acquisition of values through an accumulation of information. This element is therefore linked to the subject [98]. This aspect is seen as being stable and difficult to change because of the exposure to social media content. The informational element is related to the purpose of the exchange. It is more likely that social media users are more flexible in this regard and thus adopt or accept a point of view that is different from the one held (or not) before the social media interaction [72]. Sentiment analysis in social media combines psychological and sociological approaches to understand the building blocks of an individual’s behavior and, more importantly, their group attitude. Morphosyntactic labeling is one of the best ways to analyze feelings on social media [99]. It is a process of assigning a part of speech to each word in a sentence. This method is used to retrieve information and clarify ambiguity in what is said on networks, and it is essential in processing information. This process also allows easy classification in terms of the following:

1. Closed grammatical categories with a fixed set of words and a determined function in a language, such as pronouns, prepositions, determiners, and conjunctions

2. Closed class category composed of many words and even invented expressions; here, it is possible to find nouns, verbs, adjectives, and adverbs.

The process consists of finding borders between the implicit forms of the expressions and those belonging to the interacting group’s metaculture. Thus, sentiment analysis can be used to identify critical trends during increasing interaction on social networks [100], especially during crises.

Sentiment analysis is necessary during a crisis because it provides insight into the morale of users as well as their perceptions of the attribution dynamic. In crisis management, attribution theory is the one that consists in evaluating who is to blame for the crisis. It goes through the analysis of the threat, the determination of the initial responsibility, and the examination of the intensification factors with two elements in mind: consistency (this type of crisis has happened before) and what distinguishes the crisis from other crises [12]. Depending on the nature of the crisis, the thoughts and perceptions of victims, key players, and the public can increase the accountability of the crisis and damage the reputation of the organizations involved or affected by it. The analysis of social media sentiments in this context offers the opportunity to gauge the perception of those immersed in the heart of this crisis and those closely observing its unfolding. The position of the actors at the front of the scene has an impact on the general interpretation of the evolution of the crisis. There are close links between the attribution of responsibility in a crisis and the post-crisis reputation of an organization [79, 101, 102, 103, 104]. One could use social networks by scanning opinions and analyzing sentiments on social networks. It is therefore possible to determine how the crises were perceived by those who experienced them from the front lines.

6.2 Quantitative analysis

This section focuses on the literature that describes the use of data models and statistics to explain the complexity and the links between social media and crisis management.

6.2.1 Network analysis

SNA provides information on how people and institutions function during and after disasters and adapt to hazard settings [105]. SNA is the mapping and measurement of relationships and flows between connected people, groups, organizations, computers, URLs, and other information/knowledge entities in a crisis. The network nodes are the people and groups, whereas the links show the relationships or flow between the nodes. SNA provides a visual and mathematical analysis of human relationships [106]. Organizational network analysis allows one to X-ray an organization and reveals the organizational nervous system that connects everything. The analysis of social ties is a field that has grown in importance from establishing the sociology of information and communication techniques [56].

Network analysis allows mapping conversations on a topic to distinguish network types based on their division, density, and direction [107]. There are two types of network analyses: egos network analysis (EAN) and complete network analysis. The EAN assesses the nodes that shape the relationships between actors in the network, making it possible to highlight the operating radius, the diameter of the node and its centrality, the proximity, and the betweenness interaction [108]. While managing a crisis, this exercise allows us to gauge the scope and, therefore, the capacity of influence of information circulating in the network. One can thus determine the nature of the network from the typology of Smith et al. [46]. The centrality of a node measures its prominence or structural importance in a network. A high centrality score can indicate power, influence, control, or status. Determining which node is the most “central” can help disseminate information more quickly in a network, stop epidemics, protect a network from disruption, and identify suspected terrorists, among other things. Network visualization allows for a series of centrality measures to identify the most influential nodes in a social network.

6.2.2 Netnography

Netnography is the adaptation of ethnographic studies on the behaviors and interactions of people on social media. It allows one to map out the key stakeholders and their behaviors during a crisis. It also helps the CMT to gain a clear understanding of the digital infrastructure involved in the crisis [109]. It provides information visualization critical to essential decision making, such as managing the response to a crisis. The technique was efficiently used in the case of the 2015 earthquake in Nepal to establish platforms of communication and effective collaboration between the officials managing the response to that disaster and the stakeholders (the general population and the victims; [110]). Netnography needs to be developed before the crisis, which allows the CMT to know exactly who to talk to and when as well as to anticipate their potential reaction by updating their frequently asked questions to respond appropriately and provide relief.

6.2.3 Applications for disaster

When it comes to crisis triggered by natural disasters, accurate and quality information is critical to predict, prevent and manage the situation. To manage such a crisis, there are a series of tracked metrics, such as TTR (time to respond); MTBF (mean time before failure); MTTR (mean time to recovery, repair, respond, or resolve); MTTF (mean time to failure); and MTTA (mean time to acknowledge), which can help teams properly implement their response. For extreme events such as natural disasters, it is critical to connecting people in need and dire situations with institutions and people who are willing or able to help. Numerous web applications have helped to achieve that results whether during earthquakes, hurricanes, floods, and snow storms by successfully enabling civilians to use their smartphone to produce data, describing the scene and channeling support [111]. Social media in general and web applications, in particular, have been involved in the short-term return to normalcy (Table 2) [117, 118].

On top making facilitating the collect data, disaster management apps also render the analysis and the process of these data easier by increasing data visualization thus transforming information (raw data) into intelligence (useful information) that could be used to save lives [119].

With the increased use of mobile technology, researchers dealing with crisis management have been developing applications that could help teams foster their ability to use the metrics mentioned above, enabling them to efficiently report, track, and share information as well as interact with victims during emergency situations [8]. The main advantage of these applications is how they leverage data to integrate them into multiple mobile technology sources, thus increasing victims’ contribution to the collective knowledge of the extreme event dynamics and intervention team visualization of the situation. These applications have proven useful for critically matching the needs with available resources at the right moment. In some cases, such as the earthquake in Haiti, interventions started on-site, thanks to applications like Ushahidi, an open-source platform providing insights into events happening in near-real time [115]. Crowdsourcing applications and volunteered geographic information have been used in numerous disasters for preparedness, mitigation, and response purposes (Figure 3).

Figure 3.

Breakdown of applications used per crisis type as appeared in the reviewed literature.

6.2.4 Community response grid

Social media network analysis also clearly allows CMT to understand community response grids (CRGs). CRGs integrate mobile technologies and social media, enabling people on a disaster site to report information and respond to instructions from the intervention team, thus facilitating intervention in large-scale emergencies [120]. According to Wu et al. [121], a CRG is a geographically based sociotechnical network that helps local communities become better prepared for and more resilient to emergencies. Empowered by the Internet and mobile technologies, the system helps local communities establish multichannel emergency communication, report emergencies to officials, receive information from official and community sources, coordinate peer-to-peer assistance, provide emotional support, and build trust.

7. Conclusion

Social media is incredibly present in people’s lives globally. It plays a critical role in managing crises as well. Research has reflected on multidisciplinary approaches to demonstrate the impact of social media on an organization’s mitigation, preparedness, readiness, and response and recovery deployment during a crisis. This chapter conducted a systematic literature review assessing how researchers have increasingly used real-life case study models considering related events, precursors, and unperceived variables to understand the governance of uncertainty through social media.

Content analysis focusing on discourse, narratives, videos, and pictures used during a crisis appears to be one of the main methods in the literature to outline social media’s influence during a crisis and its use by the crisis management team. Another important method in the domain present in the literature is the network analysis. Here, researchers focus on connectivity and interactivity by analyzing the use of centrality and the importance of nodes during a crisis. The degree of centrality is the most straightforward measure of node connectivity. Sometimes it is helpful to consider degree (number of inbound links) and degree (number of outbound links) as separate measures, for example, when examining transactional data or account activity during a crisis. This method allows the crisis management team to properly channel their responses into action by providing critical information about the most influential actors and subjects. Social media increases the understanding of two crucial elements during a crisis: stakeholders’ core needs and concerns. That said, this research has its own limitations. In effect, a crisis can be considered as both danger and opportunity, and social media presents a dangerous opportunity and must be handled with care in times of crisis. With its propensity to spread fake news, misinformation, and rumors, social media can often become a double-edged sword. It can be found at the origin of a crisis and can contribute to amplifying a crisis by amplifying it on a disproportionate scale. This dangerous aspect has not been addressed in this chapter and could become the object of another research project in the future in a holistic attempt to comprehend the main links between crisis management and the use of social media.