Content area
Purpose
In social media, crisis information susceptible of generating different emotions could be spread at exponential pace via multilevel super-spreaders. This study aims to interpret the multi-level emotion propagation in natural disaster events by analyzing information diffusion capacity and emotional guiding ability of super-spreaders in different levels of hierarchy.
Design/methodology/approach
We collected 47,042 original microblogs and 120,697 forwarding data on Weibo about the “7.20 Henan Rainstorm” event for empirical analysis. Emotion analysis and emotion network analysis were used to screen emotional information and identify super-spreaders. The number of followers is considered as the basis for classifying super-spreaders into five levels.
Findings
Official media and ordinary users can become the super-spreaders with different advantages, creating a new emotion propagation environment. The number of followers becomes a valid basis for classifying the hierarchy levels of super-spreaders. The higher the level of users, the easier they are to become super-spreaders. And there is a strong correlation between the hierarchy level of super-spreaders and their role in emotion propagation.
Originality/value
This study has important significance for understanding the mode of social emotion propagation and making decisions in maintaining social harmony.
Peer review
The peer review history for this article is available at: https://publons.com/publon/10.1108/OIR-03-2024-0192.
1. Introduction
Natural disaster events are unpredictable and seriously destructive, which not only pose a threat to people’s lives and wealth, but also easily induce negative emotions, causing emotional crisis. Especially in the “post-truth” era, emotion surpasses truth as the main driving force of information dissemination (Hameleers and Yekta, 2023). Some information with strong emotions, with the fundamental goal of obtaining views, likes and comments and with the main means of exaggerating the truth and fabricating facts, incites the emotions of netizens and creates confusion in social cognition (Deng and Chau, 2021; Goldenberg and Willer, 2023). Moreover, with the help of new media platforms, emotional information is rapidly fermented through social networks, which has a serious impact on social stability (Kim and Chen, 2022). In this context, it is increasingly important to mine and understand the mechanism of emotional information dissemination in natural disaster events.
Kramer et al. (2014) and Brady et al. (2017) have found that emotions form a transmission network in the process of infection, and spread and strengthen under the influence of interpersonal relationships. Emotional information is similar to information, following step flow models (Yi et al., 2022).Users with high influence can expand the radiation range of emotional information (Naskar et al., 2020), and may also influence the emotional state of the public (Goldenberg et al., 2016; Liu and Liu, 2023) In the emotional network, users with high influence are regarded as super-spreaders, and the emotional information they publish will be widely disseminated, which greatly affects the public's cognition and emotion (Gao et al., 2023). Therefore, identifying super-spreaders has become a central key to recognizing the mechanism of emotion propagation.
In the new media era, internet audiences participate in information dissemination through different ways (Mak et al., 2024). The direct communication between participants at different levels undermines the effectiveness of the dichotomy of mass communication and interpersonal communication (O'Sullivan and Carr, 2018). The two-step flow communication theory is difficult to describe the sharing behavior of information in online networks (Mohammadi et al., 2016). Therefore, the multi-level communication theory has been proposed, which suggests that the process of information dissemination in society is multi-level, and there may be several levels of super-spreaders from the source of information to the audience (Hilbert et al., 2017). Multi-level communication theory exists not only for information dissemination, but also for emotion propagation. People may be exposed to the influence of a variety of super-spreaders in the process of emotion formation, and after more than one emotional conversion. Emotional information presents multi-level propagation characteristics.
The purpose of this study is to explain the emotion propagation in natural disaster events based on multi-level communication theory. Combining with emotion analysis and emotion network analysis methods, focusing on the information diffusion capacity and emotional guiding ability of super-spreaders, this study analyzed the flow law of emotional information in natural disaster events and the diverse leadership of super-spreaders in different levels. Through this study, we attempt to answer the following questions:
Who are the super-spreaders in the emotion network of natural disaster events?
What role do super-spreaders play in the process of emotion propagation during natural disaster events?
What are the differences in the roles and functions of super-spreaders in different levels of hierarchy?
2. Literature review
From the theory of multi-level communication, the study of emotion propagation mainly relies on social network analysis and key nodes identification, therefore the literature review of this article is organized into the following three subsections. The first subsection introduces the intrinsic meaning and significant characteristics of the multi-level communication theory. The second subsection focuses on the relationship between emotion network and social network. While the third subsection emphasizes the key role of super-spreaders identification in emotion propagation research.
2.1 Multi-level communication in social network
Multi-level communication theory developed from two-step-flow theory and emphasizes the diversification of mediating factors and links in the process of information dissemination. The two-step flow theory suggests that information tends to flow first to opinion leaders in social networks and then from opinion leaders to ordinary people (Hunt and Gruszczynski, 2024). However, the process of information dissemination is much more than two stages. From the information source to the audience, there may be amplified, filtered, interpreted and reshaped by multiple intermediary links (Chen et al., 2024; Hasell, 2021).
Especially in the new media era, the mode of information dissemination has changed significantly. People of all social classes are able to participate in the dissemination of information, and can communicate directly with each other (Laor, 2023; Li, 2022). Information chains are formed in social networks, which are influenced by opinion leaders and various intermediary factors (Tang et al., 2023; Yu et al., 2017). As a result, scholars have proposed a multi-level communication theory to more fully reflect the complex communication process of information in social networks.
In the new media era, diverse ways of expression also pose challenges to the assessment of public emotion. However, it is worth affirming that emotion propagation on social media may be influenced by both mass media and online audiences (Hu, 2022; Lu et al., 2021), which conforms to the multi-level communication theory. In the process of spreading emotional information, the mass media can usually accelerate the spread of emotion and expand the scope of information, leading some people to have the same emotional state (Laor and Lissitsa, 2022). In addition, people can also accumulate emotional energy through interaction, thus promoting emotion propagation.
Multi-level communication theory can help to accurately identify the key roles in the process of emotion propagation and analyze its role and influence. In addition, analyzing how audiences at different levels receive, understand and react to emotional information can more accurately grasp the dynamic process of emotion propagation. Therefore, the multi-level communication theory provides a theoretical framework for a comprehensive and in-depth understanding the process of information dissemination, which provides theoretical support for this paper to grasp the flow and diffusion laws of emotion propagation.
2.2 Emotion network based on social network
The background of emotion network research can be traced back to social psychology and cognitive psychology. Early research focused on exploring how people express, perceive and transmit emotions in social interaction, focusing on a variety of modes of emotional propagation such as expressions, actions and behaviors. With the development of new media technology, information plays an increasingly important role in the process of emotion propagation (Cheng et al., 2023; Luo et al., 2023). Emotion gives information emotional attributes through mass media, and then emotional information forms emotional flow through information dissemination network. Information receivers can generate the same emotions as the information senders without realizing it (Kramer et al., 2014), thus enabling the widespread of emotions across social media platforms. As a result, some of the users’ information behaviors in social networks not only convey literal meanings, but also indicate a person’s emotional attitude (Yu et al., 2019; Zhang et al., 2018).
In this context, scholars have begun to pay attention to the problem of information flow and emotion propagation on social media platforms (Baek et al., 2021; Lange and Zickfeld, 2021; Li et al., 2022). Social media platforms are convenient, fast and interactive and more and more people choose to use them to find, collect and share information (Eitan and Gazit, 2024; Triptow et al., 2024). Especially when emergencies occur, people are used to searching for the latest news, countermeasures and related developments of authoritative institutions on social media platforms (Chen and Chiu, 2022). For example, in natural disaster events, people urgently need to know the type of disaster, the scope of impact, the damage caused and other information (Dong et al., 2021; Muniz-Rodriguez et al., 2020). For epidemic events, it may involve relevant information such as the characteristics of the virus, prevention and control measures and the development trend of the epidemic (Tsao et al., 2021; Xu et al., 2024). More importantly, people can release emotions and form emotional connections and support with others through information behavior.
Social network helps people to consume information and spread emotion through their unique interactivity and information dissemination capabilities. Established studies have analyzed the impact of social network characteristics on the emotional information dissemination (Hung et al., 2023; Luo et al., 2023), explored the role of individual influence in emotional contagion (Rhee et al., 2020), explored the relationship between information behavior and emotion (Cai et al., 2022; Chung and Zeng, 2020) and discussed how to apply these knowledges to improve social well-being. The study has found that social network provides a large number of users and efficient communication channels for emotion propagation, which can quickly expand the scope of emotional information dissemination (Joseph et al., 2021). Emotional information in social networks can spread at an exponential rate, and emotions may erupt instantaneously and spread rapidly (Daou, 2021). Opinions of contacts in social network can affect each other (Li et al., 2019), and the closer the social interaction of users, the higher emotion influence between them (Wang et al., 2015). Even, some media use social networks to promote popularity and emotional impact (Laor, 2019). Therefore, modeling of emotion propagation processes should consider the changing social network.
Network theory can better explain the phenomenon of emotional information spreading in social media, so people start to build emotion network. Network nodes represent individual netizens who express their emotions on social media platforms, and connectivity can be considered as the emotional connection between users. It is worth noting that information with strong emotion is more likely to be the center of attention and receive more attention (Li et al., 2024). Emotion network and information network have different structural features (Yi et al., 2022). Therefore, building an exclusive emotion network has become an important topic for studying of emotion propagation.
2.3 Super-spreaders in the emotion network
To date, scholars have conducted many studies on opinion leaders in social network. Studies show that opinion leaders have multiple interpersonal influences, including followers influence, retweets influence and mentions influence (Turcotte et al., 2015). They rely on their own influence to “persuade” interpersonal relationships to spread information, with the function of expanding the scope of information radiation, increasing the attention and participation in the event (Gao et al., 2023; Weeks et al., 2017). Opinion leaders are often experts in specific fields with a wealth of knowledge, experience and expertise (Hassanzadeh et al., 2023). The information they publish can have a significant impact on the social ecosystem (Li et al., 2022; Naskar et al., 2020).
Super-spreaders in emotion network are similar to opinion leaders in social network. Both of them have the ability to influence people’s opinions and attitudes, and promote the diffusion of information or emotional information. Super-spreaders have strong emotional contagion, they are good at triggering and transmitting emotions and are able to inspire empathy and emotional responses in others in a short period of time (Oueslati et al., 2023). It is worth noting that super-spreaders may be authoritative media outlet or ordinary users (Yi et al., 2022). Compared with opinion leaders who rely on authoritative characteristics, the ability of super-spreaders depends on the emotional contagion of the individual.
With the support of Internet technology, the dissemination of emotional information has achieved popular communication, and the super-spreaders with high influence are scattered in various levels. In order to identify super-spreaders in emotion network, it has been proposed to quantify the social influence of users based on the number of followers, replies and reposts of the disseminators (Peng et al., 2017). In addition, methods such as social network analysis have been used to describe the unique roles played by super-spreaders in the process of emotion propagation and to explore the characteristics of super-spreaders (Du et al., 2022; Liu and Liu, 2023). Interesting conclusions have been drawn from various surveys of super-spreaders. For example, although social media platforms have weakened the information dissemination function of super-spreaders, they have not weakened the influence of super-spreaders (Choi, 2015; Weeks et al., 2017). The emotion trust distinguishes super-spreaders from others and gives them leadership position (Gao et al., 2023). Identifying super-spreaders in emotion network can help government complete positive social emotional guidance in natural disaster events.
In summary, observing super-spreaders in emotion network is an important way to study emotion propagation in natural disaster events. However, it is worth noting that under the influence of mass media, people can be exposed to multiple super-spreaders, and more than one mood shift may occur. Therefore, this study focused on analyzing the information diffusion ability and emotional guidance ability of super-spreaders in different levels of hierarchy.
3. Method
In today’s social interaction platform architecture, emotional information is hidden under the information dissemination network, and spread at exponential pace via multi-level super-spreaders. With the goal of understanding the flow pattern of emotional information and the information dissemination preferences of super-spreaders in different levels, this study combined emotion analysis and social network analysis methods to propose a reporting framework for multi-level emotion propagation in natural disaster events, as shown in Figure 1.
The reporting framework consists of five components, including data collection, emotion analysis, emotion network analysis, super-spreader identification and multi-level emotion propagation. Firstly, data collection is completed by extracting data with identified keywords from a certain time range. Secondly, emotional information is screened and observed by calculating the emotion scores of tweets. Thirdly, using the forwarding data to build an emotion network. Fourthly, based on social network analysis methods, the node centrality index is calculated to identify the super-spreaders in given emotion network. Finally, hierarchical structure is divided according to the number of followers to analyze the diverse leadership of super-spreaders in different levels.
3.1 Data collection
The data we use in this study, were collected mainly on Weibo, one of the most popular online social media platforms in China, using Python and the keyword “7.20 Henan Rainstorm” for crawling. From July 20 to July 23, 2021, China’s Henan Province was hit by extreme rainstorms, causing urban waterlogging, flooding, landslide and multiple concurrent disasters. The rainstorm has brought huge economic losses and serious life threats to the local people, and sparked extensive and ongoing discussions among Internet users. In addition, the influence index of the “7.20 Henan Rainstorm” event on social media is as high as 98.4%. It is the most discussed natural disaster event in China in recent years, which is largely representative and can provide quality data for analyzing emotional information patterns. Therefore, this study took the “7.20 Henan Rainstorm” event as an example for empirical research.
The crawled content includes release time, microblog content, user name and forwarding number. It is noteworthy that since web crawling can only return the content of the first fifty pages, this study used “hours” as the unit of search to obtain more complete data. In the end, a total of 47,042 original microblogs were collected within 6 days from 00:00 on July 20,2021 to 24:00 on July 25,2021. By the way, this study strictly adhered to ethical principles, including the principles of authenticity, impartiality, openness and transparency and data anonymization, when conducting data collection and data processing. For example, when using data, we anonymize information about ordinary users to reduce the risk of privacy leakage. In addition, we detailed the process of data collection, processing.
Pre-processing of the data prior to the experiment is divided into two parts: the first part is to perform data cleaning and screening, and the second is to perform emotional information screening with related forwarding data collection. First of all, respective microblogs were manually screened to remove errors, duplicates and irrelevant contents. This study focused on key nodes with high influence, so the data forwarded less than 50 times were also filtered out, and 976 original microblogs were obtained. Secondly, this study only analyzed emotional information, so data with emotional score of 0 were further filtered, and 702 eligible original microblogs were retained. Finally, 120,697 corresponding forwarding data were collected.
3.2 Emotion analysis
Emotion analysis is different from sentiment analysis (Park and Storey, 2023). Sentiment analysis simply categorizes emotions into positive, negative and neutral emotions, while emotion analysis provides more specific insights by categorizing emotional information into 7 types of emotions (good, surprise, sadness, happy, fear, disgust and anger) according to psychology (Yadollahi et al., 2017). Moreover, emotion analysis can be categorized into two groups, content-based analysis and diffusion-based analysis. Content-based analysis is mainly to mine the public’s views and attitudes in the comment content through emotion dictionary or machine learning methods (Aslam et al., 2023). Diffusion-based analysis focuses on the propagation characteristics, evolutionary patterns and influencing factors of emotional information. These provide a basis for the study of emotional changes in natural disaster events.
Natural disaster events are often sudden, can cause serious damages and have unpredictable development, which can easily stimulate various social emotions. First of all, when the safety of their lives and property is threatened, people will inevitably have negative stress emotions such as tension, panic, anxiety and so on (Ma et al., 2023; Xu et al., 2015). However, in the process of effective control of natural disaster events, there will still be many positive events emerging, stimulating people’s positive emotions such as pride, admiration and gratitude. Therefore, the public opinion field presented a phenomenon in which positive emotions and negative emotions are intertwined, and optimism and pessimism coexist.
Different types of emotional information are intertwined, but they may present different dissemination characteristics. The existing mature Affective Lexicon Ontology of Dalian University of Technology divides emotions into 7 categories and 21 subcategories and sets the initial intensity of emotions as five levels of 1,3,5,7 and 9 (Guo et al., 2021). Meanwhile, the degree adverbs and negative words that affect emotion analysis are excavated and included, which provides convenient and reliable help for observing emotional information.
In this study, we used the jieba word segmentation package and Affective Lexicon Ontology to calculate the emotional score of the information, so as to realize the screening and classification of emotional information. Moreover, according to the visual analysis of the trend of emotional communication in the time dimension, we observed the dynamic evolution of social emotions in the “7.20 Henan rainstorm” event. By comparing and analyzing the percentage of the number of forwarded emotional information, we illustrated the propagation efficiency of each emotion.
3.3 Emotion network analysis
Separating out emotional information and its forwarding data allows for the construction of emotion network. In emotion network, nodes represent the users who express their emotions, connections represent forwarding relationships and are considered to be emotional connections between users. Moreover, previous studies have shown that users with high influence in network society can affect the trend of emotional evolution to some extent (Cassell, 2021; Liu, 2022). In order to understand the influence of super-spreaders in different levels and to answer our research questions, we identified the super-spreaders by calculating the attribute parameters of user nodes.
Degree centrality can measure the degree of closeness between nodes and other nodes, and is the primary indicator reflecting the influence distribution and communicative ability of nodes in the network (Naskar et al., 2020). Nodes with large degree centrality are able to effectively control and influence the emotional information interaction behaviors of other nodes in the network (Zeng et al., 2021). Betweenness centrality can measure the degree of node control over emotional information (Maji et al., 2021). Large betweenness centrality indicates that the node exists on numerous linkage paths between two nodes and assumes the role of a mediator in the emotion propagation (Chen et al., 2012). Closeness centrality is an indicator used to measure the independence of nodes in emotion propagation (Luo et al., 2013). The greater the closeness centrality of a node, the closer it is to other nodes and the more difficult it is to be “controlled” by other nodes (Krnc and Skrekovski, 2020). Nodes with high closeness centrality will play the role of gossiper in the center of the emotion network and actively transmit emotional information.
Degree centrality, betweenness centrality and closeness centrality reveal the influential role of nodes from different perspectives, which provides a measurement basis for explaining the emotional guidance role of super-spreaders (Asim et al., 2019). Considering the different focuses of the indicators, this study first screened the users with high degree centrality as important super-spreaders. However, users with high degree centrality are not necessarily active emotional information transmitter. Therefore, betweenness centrality and closeness centrality were used as a further screening condition.
3.4 Division of hierarchical structure
The development of new media supported by emerging technology has broken the information monopoly of traditional media, and the participation of network audiences has created a new environment for emotional information dissemination. It is becoming common that ordinary users have more followers than official media and are followed and recognized by more and more people. The original social class is overthrown, and people elect their own leaders. In order to clarify the hierarchy of the super-spreaders and to answer the question of the differences in roles and functions on different levels of super-spreaders, this study categorized the users into five levels according to the number of followers from high to low. Based on this, the emotion propagation environment was divided into five levels (see Figure 2).
Users of the first level have at least 10 million or even more than 100 million followers, who are regarded as the head users in the emotion propagation environment. They tend to maintain a high frequency of emotional information release and are always active in the emotional flow. The users of the second and third level are often the main supporting force for emotion propagation. They tend to reach more users in different circles and interact more frequently with other users. Users with less than 10,000 followers are considered tail users. They have a large group, but are rarely active in the emotional flow, they often need to be infected and guided.
By dividing the structure into hierarchical levels, this study described in detail the diverse leadership of super-spreaders. Specifically, this study examined the correspondence between users' emotional leadership and interpersonal skills by observing the hierarchical structure of super-spreaders and comparing the hierarchical levels of super-spreaders and the roles they play.
4. Results
4.1 Emotional content and diffusion analysis
Based on the emotional classification results of the original microblogs, this study analyzed time dimension changes of social emotions. This provides a reference for exploring the emotional guidance role of super-spreaders. The results show that a positive and optimistic emotional atmosphere was quickly formed during the “7.20 Henan Rainstorm” event, with “good” reaching its peak on July 21, followed by “happy” (see Figure 3). “Good” information is steadily ranked first in the number of daily emotional information, and its trend change is almost consistent with the change of overall emotional information, indicating that the emotional atmosphere was dominated by “good” emotion. To a certain extent, this has adjusted the public’s pessimistic cognition, and inhibited the spread of negative emotions. This is evidenced by the fluctuating decline in the number of “disgust” information after reaching maximum value on July 20 and the continued low number of “sadness”, “fear” and “surprise”.
The statistical analysis of the number of emotional information and forwarding data is shown in Figure 4. The proportion of emotional information in “7.20 Henan Rainstorm” event from high to low is as follows: “good” (75.2%), “happy” (13.5%), “disgust” (6.4%), “fear” (2.3%), “sadness” (2.1%), “surprise” (0.3%), “anger” (0.1%). The proportion of “good” and “happy” information is as high as 88.7%, far more than the sum of other emotional information, almost forming a “monopoly”. In addition, the proportion of emotions has changed due to people’s choice of emotional information during the forwarding process. Obviously, “good” information is still occupying the first place, with 82.3% of the forwarding data for the same emotional information. However, the proportion of “fear” in forwarding data is higher than that in emotional information. On the contrary, the proportion of “disgust” and “happy” information in forwarding data has decreased.
In general, “good” information was most likely to spread widely in the “7.20 Henan Rainstorm” event. Previous studies have also shown that “good” is the most transmitted emotion in Weibo during the COVID-19 pandemic (Yi et al., 2022). It can be considered that people are more likely to search for and receive information with positive emotions in the face of unavoidable crisis events and their enormous losses. A large number of positive emotional information on social networks effectively hedges negative emotions and quickly forms an emotional atmosphere with “good” emotion as the absolute mainstream.
4.2 Emotion network of “7·20 Henan Rainstorm”
By using the forwarding data, the emotion network is abstracted into a graph. The node represents the user, the connection represents the forwarding relationship, and also represents an emotional identity. The color of the connection represents the emotional type. According to the principle that nodes with forwarding relationship are close, nodes without forwarding relationship are mutually exclusive and nodes with larger connection degree are larger, the emotion network visualization of “7.20 Henan Rainstorm” event was realized on Gephi (see Figure 5).
As shown, the emotion network of “7·20 Henan Rainstorm” is generally characterized by multiple cores and intertwined emotions. On the one hand, the nodes in the network are concentrated around cores, but the whole is decentralized. Specifically, before the emotional information reaches the vast majority of users, it flows to some users similar to “leaders”, making the network multi-center radial. On the other hand, many types of emotions emerged during the “7.20 Henan Rainstorm”, but the whole was dominated by the “good” and “happy” emotions released by a large number of core nodes.
Core nodes can change the attitudes, preferences and behaviors of individuals and even lead the social climate (Cassell, 2021; Liu, 2022). We took several representative core nodes as examples for profiling, and the results are shown in Figure 6. Most of the core nodes in the emotion network, such as Person Account 1 or CCTV NEWS, tend to spread only “good” or “good” and “happy” information. Some core nodes released “fear” and “sadness” information, but neither emotion was widely disseminated. Although the People’s Daily has more followers, but it released too many types of emotional information, creating an opportunity for emotional migration for the audience, which has an impact on the spread of “sadness” information. Nodes releasing “fear” information tend to release only one emotion information, but the number of nodes is too small to cause widespread dissemination. Therefore, as a leader independently selected by people, the core node has greatly affected the dissemination of emotional information.
4.3 User composition of super-spreaders
By calculating the centrality parameter of the nodes, 25 super-spreaders in the emotion network were selected. Meanwhile, in order to verify the influence of different mediators on emotion propagation, we conducted statistics and observations on the user composition of super-spreaders (the results are shown in Table 1).
Both official media and ordinary users participated in the emotion propagation, and became the super-spreaders with different advantages. Table 1 shows that the super-spreaders include 14 Blue V accounts representing official organization, 4 Gold V accounts representing hyperactive users, 3 Orange V accounts representing active users and 4 uncertified users. The official media remains the main component of the super-spreaders, who occupies more than half of the seats. However, the ordinary users are starting to make an impact and are no longer being bystanders in the emotion propagation. There are numerous super-spreaders, allowing people to access emotional information from different sources, or even receive different emotional information at the same time.
The top-ranked CCTV NEWS and other users open a significant numerical gap in degree centrality. As the official account of the Central Radio and Television Station, CCTV NEWS has strong authority and credibility, and it is easy for people to believe the information it publishes and be infected by the emotions it expresses. Similarly, People’s Daily, China News and 11 others have also gained emotional recognition by taking advantage of the influence of official media. The official media relies on its influence and the emotional trust of the population to occupy a central position (Liu, 2022; Sun et al., 2022; Tang et al., 2023). This study also demonstrates this phenomenon.
In the era of new media, ordinary users can also become leaders in emotion propagation. The data shows that the degree centrality of unauthenticated Person Account 4 is as high as 2,800, and the degree centrality of Person Account 1 is even greater than that of the People's Daily, which has 151 million followers. Therefore, active users outside the official media are also propagation force that should not be ignored in natural disaster events. They tend to get the emotional support of the masses with more relatable contents, more easily understood expressions, and more active interactions, thus becoming super-spreaders with tons of emotional connections.
Super-spreaders with high degree centrality bear the responsibility of information release, and the super-spreaders with high betweenness centrality plays the role of information transmitter. The data show that the betweenness centrality of the vast majority of super-spreaders with high degree centrality is 0. It illustrates that emotional information was mostly a one-way flow, with poor interaction between nodes. Therefore, separate super-spreaders are unable to control the trend of emotional propagation, and emotion propagation requires the participation of both official media and ordinary users.
4.4 Hierarchical structure of super-spreaders
In the “7·20 Henan Rainstorm” event, users from five levels participated together, providing rich subjects for emotion propagation (see Table 2). On the whole, emotional information with a certain amount of forwarding was mostly released by users with a follower base. Waist users with between 10,000 and 10,000,000 followers were the main disseminators of emotional information in the “7·20 Henan Rainstorm” event. Nowadays, people are accustomed to using new media platforms to connect with others, and make use of the new type of “neighbor” relationship to infect emotions and expand the scope of emotional information dissemination. Users with a large number of followers have the advantage of emotional guiding ability. Therefore, the number of followers can be a valid basis for categorizing the hierarchical levels of the emotion propagation environment.
In order to illustrate the diverse leadership of super-spreaders, this study counted and analyzed the hierarchy level and role of super-spreaders (the results are shown in Table 3). Observing the hierarchical structure of the super-spreaders, it is found that it basically follows the rule that the higher the level, the more the number of super-spreaders. Among the five levels we re-divided, the fifth level did not appear as super-spreader, and the first to fourth level appeared respectively 10, 5, 6 and 4 as super-spreaders. Users in the fifth level have a very small number of followers. Although they released some emotional information, they lack the interpersonal relationship basis to promote the spread of emotional information, making it difficult to complete successfully emotion propagation. On the contrary, users in the first level have a large number of followers, so they can complete emotional information diffusion and emotional contagion in a short time, becoming the super-spreaders in emotion propagation. The number of followers is an important indicator of a user's interpersonal skills, while followership is an important way to engage in emotional contagion. Therefore, there is a huge correlation between users’ emotional guiding ability and interpersonal skills.
Observing the role of the super-spreaders, we find an interesting correspondence between the hierarchy level of super-spreaders and their role in emotion propagation. The super-spreaders who assume the role of information transmitters are often users in the third or fourth level, while the super-spreaders who assume the role of information publishers are mostly users coming from the first and second level. In particular, the super-spreaders of the first level only played the role of publisher in the “7.20 Henan Rainstorm” event. Their large number of followers makes them become central in the emotion network simply by releasing information. The interpersonal skills reflected by hierarchy fully support the role of users in the emotional network.
5. Discussion
In natural disaster events, a large amount of public opinion information (including objective truth and emotional expression) emerges on social media platforms and quickly forms an information dissemination network (Li et al., 2023; Ma et al., 2023; Zhang et al., 2024). The influence of key nodes in the network determines the scope and rate of information dissemination (Krnc and Skrekovski, 2020; Sun et al., 2022). Moreover, the influence of nodes in different positions of the network are not the same, and their roles in information dissemination are also different (Bartal and Jagodnik, 2021; Wang et al., 2015). Therefore, when explaining the emotion propagation in natural disaster events, we focused on the information diffusion capacity and emotion guiding ability of super-spreaders from all levels. Through our analysis, the following points are proposed.
Firstly, similar to opinion leaders in social network, super-spreaders in emotion network have the ability to regulate the social emotional atmosphere. Although super-spreaders belong to a minority group in the process of emotional information dissemination, the emotion network generally presents a radial shape with super-spreaders as the core. There is strong correlation between the trend of emotion propagation and characteristics of the emotional information released by super-spreaders. During natural disaster events, people are likely to express negative emotions such as panic and anger (Dong et al., 2021). This study found that when the super-spreaders show a positive attitude of overcoming difficulties, “good” information is most likely to spread widely and “good” will become the mainstream social emotion.
Secondly, official media and ordinary users have become central in emotion propagation with different propagation advantages, and the environment of emotional information dissemination has been reshaped. For example, official media such as NEWS CHINA, People’s Daily, China News can rely on their own influence and people’s emotional trust to occupy the core position in the emotion network. Ordinary users such as Person Account 11 can also become key nodes in the emotion network by virtue of their positive and active communication interactions and high betweenness centrality. Some studies focused solely on the ability of the official media to lead (Liu, 2022; Tang et al., 2023), but this study shows that official media is no longer the only way for people to get information, reminding us to rethink the social structure on social media platforms.
Finally, the number of followers becomes a valid basis for classifying the hierarchical levels of super-spreaders. Previous studies have mostly analyzed the composition of opinion leaders based on account type (Gao et al., 2023), but many ordinary users have gained more attention and recognition than the official media. We categorized the users into five levels according to the number of followers from high to low, and the results show that the interpersonal skills reflected by hierarchy is not only strongly correlated with the user’s emotional guiding ability, but also explains that the user plays an important role in the emotional network.
Users with a large number of followers are able to complete the information dissemination and emotional contagion in a short period of time, becoming super-spreaders in the process of emotion propagation. The opposite is true for users with a small number of followers. For example, CCTV NEWS and Henan Meteorology are both official media. The information released by CCTV NEWS of the first level is forwarded 8,952 times, while the information released by Henan Meteorology of the third level is only forwarded 36 times. Meanwhile, users of the top level bear the responsibility of emotional information publishers, while users with fewer followers are often responsible for emotional information transmission. For example, the super-spreaders in the first level were all publishers, and five of the seven transmitters were users in the third or fourth level.
This study has important significance for understanding the mode of social emotion propagation and making decision in maintaining social harmony. Understanding the hierarchical structure of the emotional information dissemination environment helps the authority and decision makers rationalize human resources, maximize the efficiency of personnel use and improve the efficiency of emotional information monitoring and public opinion management in natural disaster events. Specifically, the authority and decision makers can subdivide the structure of participants in a crisis event, focusing on waist users. Even, some waist users can be trained as unofficial opinion leaders to use their advantages of frequent interaction to accelerate the spread of positive emotional information. Taking the official media at the center of the emotional network as the leader, together with the active dissemination of ordinary users, it is possible to create a “highway” for the emotion propagation. This helps the authority to stimulate positive emotions in a short period of time, and inhibit the spread of negative emotions.
6. Limitations and future research
There are some limitations to this study. Firstly, although we collected data on an hourly basis, Weibo’s web crawling rules made it difficult to obtain complete sample of microblogs. Moreover, under the government’s policy, some negative emotional information was blocked or deleted (Cassell, 2021), which may also be another possible reason why negative emotions in this study have not been widely disseminated. Secondly, the way people express their emotions in today’s social media is no longer limited to words, and audio content has become the primary means for people to spread emotional information with the advantages of more intuitive and higher dissemination rate. In the future, data collection will be considered for different media platforms, integrating text, video, pictures and other forms of information to enrich data resources and further generalize and enhance the scientific value of the research.
This research was funded by the “National Social Science Foundation of China”, grant number “22BXW038”.
The data that support the findings of this study are available in OpenICPSR at https://doi.org/10.3886/E199202V1.
In the interest of transparency, data sharing and reproducibility, the author(s) of this article have made the data underlying their research openly available. It can be accessed by following the link here: OpenICPSR at https://doi.org/10.3886/E199202V1
Figure 1
Reporting framework for multi-level emotion propagation in natural disaster events
[Figure omitted. See PDF]
Figure 2
Hierarchical structure in the emotion propagation environment
[Figure omitted. See PDF]
Figure 3
Emotion evolution trend of the “7·20 Henan Rainstorm”
[Figure omitted. See PDF]
Figure 4
Percentage of the number of emotional information and forwarding data
[Figure omitted. See PDF]
Figure 5
Emotion network of “7·20 Henan Rainstorm” with seven emotion types
[Figure omitted. See PDF]
Figure 6
Emotion networks of some typical nodes
[Figure omitted. See PDF]
Table 1
Super-spreaders in the “7·20 Henan Rainstorm” event
| User | Type | Degree centrality | Betweenness centrality | Closeness centrality |
|---|---|---|---|---|
| CCTV NEWS | Blue V | 8,952 | 0.0 | 1.000 |
| Person Account 1 | Gold V | 6,608 | 0.0 | 0.873 |
| NEWS CHINA | Blue V | 5,103 | 0.0 | 0.995 |
| People’s Daily | Blue V | 4,529 | 0.0 | 0.984 |
| Person Account 2 | Orange V | 4,169 | 0.0 | 0.985 |
| Person Account 3 | Gold V | 3,776 | 0.0 | 1.000 |
| China News | Blue V | 3,544 | 0.0 | 1.000 |
| Liu Yaowen’s support club | Blue V | 3,075 | 0.0 | 0.991 |
| Person Account 4 | Unverified | 2,800 | 0.0 | 1.000 |
| Dahe Daily | Blue V | 1922 | 0.0 | 1.000 |
| Henan Communist Youth League | Blue V | 1,560 | 0.0 | 0.755 |
| Person Account 5 | Gold V | 1,490 | 1,505.0 | 0.971 |
| Song Yaxuan’s support club | Blue V | 1,406 | 0.0 | 1.000 |
| Guancha Network | Blue V | 1,249 | 0.0 | 1.000 |
| Person Account 6 | Orange V | 1,177 | 0.0 | 0.634 |
| Person Account 7 | Unverified | 1,110 | 0.0 | 0.963 |
| Person Account 8 | Unverified | 1,090 | 0.0 | 1.000 |
| Phoenix Weekly | Blue V | 1,085 | 0.0 | 1.000 |
| Guangzhou Daily | Blue V | 1,054 | 0.0 | 1.000 |
| National Meteorological Center | Blue V | 990 | 2096.0 | 0.943 |
| Henan Meteorology | Blue V | 36 | 1,083.0 | 0.493 |
| Person Account 9 | Gold V | 580 | 569.5 | 1.000 |
| Person Account 10 | Orange V | 501 | 498.0 | 1.000 |
| Person Account 11 | Unverified | 396 | 388.5 | 1.000 |
| China Weather | Blue V | 384 | 383.0 | 1.000 |
Source(s): The data in this table were based on authors’ collection on Weibo, and analyzed using Gephi for centrality metrics
Table 2
Cross table of hierarchy levels and emotional information
| Anger | Disgust | Fear | Sadness | Surprise | Good | Happy | Sum | |
|---|---|---|---|---|---|---|---|---|
| Fans > 10,000,000 | 0 | 7 | 5 | 9 | 0 | 90 | 6 | 117 |
| 1,000,000 < Fans ≤ 10,000,000 | 0 | 21 | 4 | 2 | 2 | 167 | 29 | 225 |
| 10,000 < Fans ≤ 1,000,000 | 1 | 12 | 5 | 2 | 0 | 154 | 34 | 208 |
| 100 < Fans ≤ 10,000 | 0 | 4 | 2 | 1 | 0 | 86 | 16 | 109 |
| 0 ≤ Fans ≤ 100 | 0 | 1 | 0 | 1 | 0 | 31 | 10 | 43 |
Source(s): The data in this table were based on authors’ collection on Weibo
Table 3
Hierarchy level and role of super-spreaders
| User | Level | Role | User | Level | Role |
|---|---|---|---|---|---|
| CCTV NEWS | 1 | Publisher | Guancha Network | 1 | Publisher |
| Person Account 1 | 1 | Publisher | Person Account 6 | 3 | Publisher |
| NEWS CHINA | 1 | Publisher | Person Account 7 | 4 | Publisher |
| People’s Daily | 1 | Publisher | Person Account 8 | 4 | Publisher |
| Person Account 2 | 4 | Publisher | Phoenix Weekly | 1 | Publisher |
| Person Account 3 | 2 | Publisher | Guangzhou Daily | 1 | Publisher |
| China News | 1 | Publisher | National Meteorological Center | 2 | Publisher Transmitter |
| Liu Yaowen’s support club | 2 | Publisher | Henan Meteorology | 3 | Transmitter |
| Person Account 4 | 3 | Publisher | Person Account 9 | 3 | Transmitter |
| Dahe Daily | 1 | Publisher | Person Account 10 | 3 | Transmitter |
| Henan Communist Youth League | 1 | Publisher | Person Account 11 | 4 | Transmitter |
| Person Account 5 | 3 | Publisher | China Weather | 2 | Transmitter |
| Song Yaxuan’s support club | 2 | Publisher |
Source(s): The data in this table were based on the results of this study
© Emerald Publishing Limited.