Abstract
With the popularity of mobile terminals and social media increasing, misinformation about science has increased in China. To understand the nature of the popularity of scientific rumors, we analyzed 206 typical cases released by four authoritative platforms in China from 2010 to 2020. Content analysis revealed that the majority of scientific rumors are related to health and safety (76.8%), use a visual format (61.2%), are published on social media (62.2%), and provide more than three narrative elements (78.2%). In addition, rumors from unidentified netizens’ claims or homemade experiments are the most common (35.9%), followed by highly credible sources, such as expert assertions (20.9%) or scientific research results (19.4%). A further qualitative comparative analysis indicated that a combination of details and fear-mongering are significant conditions that make rumors receive significant attention. Visual presentation also plays an important role, while state media and the presence of scientific terminology have a weak effect.
Plain Language Summary
With the popularity of mobile terminals and social media increasing, misinformation about science has increased in China. To understand the nature of the popularity of scientific rumors, we analyzed 206 typical cases released by four authoritative platforms in China from 2010 to 2020. Content analysis revealed that the majority of scientific rumors are related to health and safety (76.8%), use a visual format (61.2%), are published on social media (62.2%), and provide more than three narrative elements (78.2%). In addition, rumors from unidentified netizens’ claims or homemade experiments are the most common (35.9%), followed by highly credible sources, such as expert assertions (20.9%) or scientific research results (19.4%). A further qualitative comparative analysis indicated that a combination of details and fear-mongering are significant conditions that make rumors receive significant attention. Visual presentation also plays an important role, while state media and the presence of scientific terminology have a weak effect. Limitations include lack of direct surveys of people’s perceptions, the possibility of incomplete influencing factors, and the fact that the outcome variable is influenced by other factors over a long time span.
Introduction
Information related to science has received extensive public attention, especially during the COVID-19 pandemic. However, false or misleading information has spread at an unprecedented speed, forming the first true social media “infodemic” (Karen & Tanya, 2020). Vosoughi et al. (2018) found that, on social media, fake news proliferates significantly further, faster, deeper, and more broadly than the truth. Facebook and Twitter users are both more inclined to share false rather than evidence-based information (Bessi et al., 2015; Pulido et al., 2020). However, social media platforms are not effective for refuting rumors. A. Y. Chua and Banerjee (2017) discovered that Facebook users barely reacted when rumors are proven false. Starbird et al. (2014) found that some corrections to rumors emerged in their study, but these corrections were muted compared to the spread of rumors. Due to the high potential for harm and easy dissemination and how difficult it can be to effectively counter rumors today, it is important to conduct further research on rumors and how they spread.
In China, on the one hand, the government has always promoted the slogan “Don’t create, believe or spread rumors” and even implemented a series of laws and regulations in recent years to specify the standards of punishment. However, rumors are emerging in an endless stream. A single scientific rumor is 33.302 times more influential than a social rumor (Zeng et al., 2021). On the other hand, the Chinese public has low motivation in verifying rumors. Even among those who tried to verify information, most relied exclusively on heuristic processing cues, such as source credibility, linguistic and visual cues, and intuition (Zou & Tang, 2021). Moreover, although several researchers have investigated the spread of scientific rumors in the US and European settings, very few have examined their spread in non-western contexts (Ge et al., 2021; X. Wang & Song, 2020). Therefore, to better understand Chinese scientific rumors, this study answers the following two research questions:
What are the characteristics of scientific rumors in China?
What factors influence the prevalence of scientific rumors?
Literature Review
Definition of Rumor
In developing the concept of rumors, Knapp (1944) defined a rumor as a proposition for belief, concerning a topical reference disseminated without official verification. He identified three types of rumors—pipe dreams (wishes and desires), bogies (fears), and wedge drivers (rumors dividing people and undermining relationships). Allport and Postman (1947) defined it as a proposition for belief without secure standards of evidence. DiFonzo and Bordia (2007) argued that rumors are “unverified and instrumentally relevant information statements in circulation that arise in contexts of ambiguity, danger, or potential threat and that function helps people make sense of and manage risk.” In China, the corresponding word for rumor is 谣言(yaoyan). The word “yao” originally referred to folk songs but gradually came to be used to describe gossip unsupported by fact (Kang, 2018; Zou & Tang, 2021). An article published in Public Understanding of Science investigated the definition of rumors among the public in China through interviews. They found that people typically defined rumors in terms of one or two of the following three characteristics: non-factual information, information unsanctioned by the government, and information causing panic (Zou & Tang, 2021).
Although the precise definition of rumor varies, rumors are generally unverified (Zubiaga et al., 2019). Thus, in some cases, the rumors can be true. As Fine (2007) stated, “Falsity is not a defining characteristic of rumor. Rumors may either be true or false.” In China, “piyao,” an official rumor-debunking platform (https://www.piyao.org.cn/2020yqpy/) sets the record straight on every rumor collected by rating it as “true,”“false,” or “undetermined” with the help of professionals, researchers, and reporters. Although most of them are false rumors, there are still some that are true (Zhao et al., 2022). For example, in December 2019, Dr. Li, an ophthalmologist at Wuhan Central Hospital, warned friends and colleagues about the possible identification of a novel and deadly virus in a private chat group on WeChat. Later, he was summoned by the local police for making a rumor online about the unconfirmed SARS outbreak (C. Zhang et al., 2022). However, the subsequent COVID-19 outbreak confirmed that this was a true rumor.
Shin et al. (2018) compared disinformation, misinformation, and rumor in terms of both falsity and motivation. Both misinformation and disinformation refer to untrue information. However, misinformation is agnostic regarding the motivation for falsehood, while disinformation assumes that inaccuracy stems from deliberate intention. A rumor is defined as information that is usually not confirmed, where both the falsity and motivation of the rumor are unknown. Therefore, a rumor may turn out to be true, even when it was not supported by concrete evidence at the time of circulation (Shin et al., 2018). This study uses the definitions of rumor given in the study by Shin et al. (2018). A scientific rumor is a type of rumor whose content is related to science and scientific knowledge (D. C. He, 2017; Yang & Shi, 2011). Buckner (1965) argued that “whether a rumor is truthful or untruthful, although it has been a matter of debate, is unimportant in studying its transmission.”
Rumor Transmission
Why do rumors arise? Social psychologists argue that rumors arise in the context of ambiguity, that is, when the meaning of a situation is not readily apparent, or in the context of potential threat, that is, when people feel an acute need for security (Zeng et al., 2021). Knapp (1944) argued that a rumor may in the beginning arise out of mild curiosity or suspicion, and it spontaneously reflects public opinion. Others view rumor as a form of collective problem-solving at a time of uncertainty (Fisher, 1998). Rumors fulfill specific cognitive needs by allowing the public to make sense of an ambiguous situation together and help people cope with emotions, such as fear, anxiety, and uncertainty (Rosnow, 1988). Rosnow (1980) claimed that rumors originate in wants, needs, or expectations stimulated by anxiety-producing events. Some surveys have reached similar conclusions. A survey found that the lack of a reliable news resource was considered the most common cause of rumors (63.6%) (Banakar et al., 2021). Ariel et al. (2022) discovered that individual drives, shaped by personal needs and degree of negative feelings, were the leading factors behind rumormongering. Generally, rumors arise in situations that are personally relevant but ambiguous or cognitively unclear and when credible explanations are not available from traditional sources, such as mass media, government agencies, or corporate management (Bordia & Difonzo, 2004).
Regarding rumor transmission, Allport and Postman (1947) proposed that the importance and ambiguity of rumor were related. Researchers have also referenced the influence of public critical thinking ability, anxiety level, source, and trust (Chorus, 1953; Jaeger et al., 1980). Rosnow (1991) summarized four factors that impact the spread of rumors—personal anxiety, general uncertainty, credulity, and outcome-relevant involvement. The meta-analysis reflected the important role of anxiety in the spread of rumors and the minimal effect of uncertainty (Rosnow, 1991). J. Wu and Ma (2015) noted that, after Rosnow, most western scholars have focused on further testing or refining the above factors. For example, O. Oh et al. (2013) found that the ambiguity of the source was the most important factor in disseminating rumors, while content ambiguity was not found to affect rumors. Chinese scholars have added other factors, such as attention and abnormality (Hu, 2000); information asymmetry (Kuang & Guo, 2012); and news value, media, and control (J. Wu & Ma, 2015). C. F. Wang (2010) identified the political environment and communications network in China as prominent factors in the spread of rumors. These Chinese studies are mainly based on literature reviews or case studies. F. Zhang and Zhang (2017) argued that most previous studies in China have been limited to empirical summaries and theoretical generalizations. In general, there are a few empirical research results (Gao et al., 2023).
Regarding scientific rumors, health rumors have received as much attention as those related to the public interest. Many studies have analyzed scientific rumors based on the above-mentioned factors. In addition, other factors are related to the information level, such as the length of texts, the presence of images (A. Y. Chua et al., 2016), the sentiments of the rumor (Na et al., 2018), presence of URLs (Tanaka et al., 2014), and hashtags (Zhou et al., 2018). Factors related to the individual rumor recipient, such as original beliefs (DiFonzo et al., 2012), scientific literacy (L. He et al., 2021; W. He et al., 2021; H. J. Oh & Lee, 2019), and other demographic variables (Pan et al., 2021) have also been discussed. Moran et al. (2016) found that stories or personal testimonials can enhance persuasion. Studies have also considered the influence of social media and rumor corrections. Social media has been demonstrated to be effective in correcting misinformation (Bode & Vraga, 2018; Vraga & Bode, 2018). In terms of social relations, Seah and Weimann (2020) found that a stronger altruistic motive leads to a higher willingness to forward health rumors. In general, rumors spread because of many factors.
Most studies that have investigated the influencing factors used experimental research, empirical summary, theoretical generalization, case study, and structural equation modeling. They often test the effect of a single factor on rumors rather than a combination of factors.
QCA in Communication
Qualitative comparative analysis (QCA) was originally developed by Charles Ragin in 1987, whose toolbox has expanded from basic, dichotomous QCA (crisp-set QCA—csQCA) to fuzzy-set QCA (fsQCA) and multi-value QCA (mvQCA). QCA was originally developed to analyze small- to medium-sized datasets (Fiss, 2007). However, researchers are increasingly using QCA to analyze large datasets (Meuer & Rupietta, 2017; J. Wu & Zhou, 2022).
Ragin’s whole intention was to develop an original “synthetic strategy” as a middle way between case- (or “qualitative”) and variable-oriented (or “quantitative”) approaches, which would “integrate the best features of the case-oriented approach with the best features of the variable-oriented approach” (Ragin, 1987). Based on set theory and Boolean algebra, QCA allows the discovery of a combination of conditions, leading to a concrete outcome (Rabadán et al., 2020). More precisely, QCA develops the concept of causality referred to as “multiple conjunctural causation” (Rihoux & Ragin, 2008). Specifically, a combination of conditions (i.e., a “configuration” in QCA terminology) generates an outcome (AB → Y). Several different combinations of conditions may produce the same outcome (AB or CD → Y) (Rihoux & Ragin, 2008).
In the late 1980s and early 1990s, QCA was mostly developed for application in political science (comparative politics) and historical sociology (e.g., welfare state studies) (Rihoux & Ragin, 2008). Until about 2003, QCA use remained confined to a relatively narrow niche; since then, its use has grown rapidly, especially since 2008 (Rihoux, 2013; Rihoux, Alamos-Concha et al., 2013). As of 2011, csQCA was the most frequently used technique, and political science, sociology, and management are the core disciplines of application. A recent review found that comparative politics accounts for the largest number, followed by business and economics and sociology (Rihoux, Alamos-Concha et al., 2013). Currently, 54% of articles using the QCA methodology are published in these three areas (Roig-Tierno et al., 2017). Kim et al. (2016) argued that while the use of QCA in journalism studies is rare, our use of the method stems from the tradition of using it to compare social systems that are better described with qualitative variables.
In recent years, an increasing number of studies in media and communication science have used QCA to identify causal relationships between conditions and outcomes (Downey & Stanyer, 2010; Humprecht & Büchel, 2013). However, at the international level, QCA methods have not been often used in communication studies, especially rumor transmission. Specifically, these articles focus more on causal configurations of the following topics: political communication on social media (e.g., political parties campaigning and pop communication of key political leaders) (Ceccobelli, 2019; Spierings & Jacobs, 2019), generation or reversal of online public opinion (Li & Wu, 2022; M. Wang & Sun, 2021), satisfaction or widespread dissemination of media content in SNS (Cerdá-Mansilla et al., 2021; Mattke et al., 2020; I. O. Pappas et al., 2020), and sustainability or reporting of news websites (e.g., the occurrence of high diversity in news reporting) (Humprecht & Büchel, 2013; Kim et al., 2016).
In terms of rumor-related international articles, Neo (2022) studied the securitization of fake news through a fsQCA analysis and found a strong association between it and conditions of low media freedom, low democracy, good economic performance, political turbulence, and election proximity. X. Zhang et al. (2022) evaluated the combined impact of psychological motivation on the spread of a brand rumor. The results revealed that emotional stimulation is a key component in spreading rumors, and altruism and relationship management motivation can coexist at times (X. Zhang et al., 2022). In China, some studies focused on rumors in emergencies (Han & Fan, 2020); some addressed rumors in the health field (H. R. Wu & Luo, 2019); and some analyzed the negative effect of rumor communication (Han & Fan, 2020). The sample size examined in these studies is often small, with most not exceeding 20. In general, there is a lack of causal configuration analysis of rumor dissemination through QCA, let alone scientific rumor.
Data and Empirical Strategy
Samples
In China, four influential platforms have published annual top 10 lists of scientific rumors nearly throughout the last decade (Table 1), which are China Science Communication (科普中国), Guokr (果壳), WeChat Rumor Filter (微信谣言过滤器), and Tadpole Stave (蝌蚪五线谱). Therefore, they were selected as target research samples for this study. Each rumor selected for the study was ultimately refuted by experts, and the relevant corrections are publicly available online. The samples of rumors for this study were collected in April 2021.
Scientific Rumor Lists Collected From Each Platform.
Note. (1) Some years are missing because some platforms did not publish lists in that year. (2) The URLs of the related ranking lists are as follows: China Science Communication (https://piyao.kepuchina.cn/hotlist/hotlist), Guokr (https://www.guokr.com/scientific/channel/fact/), Rumor Filter is an official account based on the Wechat APP rather than a website, more lists could be searched in Wechat, Tadpole Stave (http://www.kedo.net.cn/news/kxpy/index_2.shtml)
China Science Communication is a political online platform established by the China Association for Science and Technology (CAST); it published eight annual lists of rumors before April 2021. Guokr is the most influential folk science popularization platform in China (WKEPU, 2019). It is known for making engaging and attractive public communication of science in the new media age. Guokr has released eight annual lists of rumors to date. WeChat Rumor Filter, an official rumor-counting platform founded by WeChat (the app with the most users in China), has published five annual lists. Finally, Tadpole Stave is a large public welfare science portal invested in and constructed by the Beijing municipal government and maintained by the Beijing Science and Technology Association. It has posted six annual lists of scientific rumors. The specific years of the lists collected from each platform are included in Table 1. The annual lists published by the platforms above usually include the top 10 scientific rumors per year, except for WeChat Rumor Filter’s list, which featured up to 30 rumors in some years (not all are scientific rumors). Generally, the lists differ from platform to platform; however, some rumors have been duplicated across platforms. The screening criteria and process for each platform are not publicly available; this paper applies a common set of criteria to reclassify the effect of rumor, as detailed in Section 2.2. Using the selection rules above, a total of 206 scientific rumors were collected (Supplemental Table S1) after eliminating duplicate rumors.
Methods
Content Analysis
In this paper, two different methods were used. This paper first describes and summarizes the basic environment and characteristics of scientific rumors in China using content analysis. Specifically, there are six variables, namely theme, presentation format, source, original communication platform, number of narrative elements, and number of occurrences. Source refers to the source or basis of the information, including expert assertions, scientific research, gossip, news reports, commercial advertisement, and non-expert claims or homemade experiments. Rumors can spread across multiple platforms, and this article only analyzes the original platforms on which they appear. Some of the content of the rumor itself mentions the original source, while others show the platform features, such as screenshots of WeChat chats; we select the platform that posted the earliest. The detailed categories of each variable and their respective proportions are presented in Supplemental Table S3. This paper then explores the conditions that influence the impact of rumor communication using the csQCA.
CsQCA
In csQCA, every case is assigned one of two possible membership scores in each set included in a study: “1” (membership in the set) or “0” (non-membership in the set); an object is either in or out of the set. However, fsQCA allows researchers to calibrate partial membership in sets using values in the interval between “0” and “1” without abandoning core set theoretic principles, such as the sub-set relation (Rihoux & Ragin, 2008). For this study, all the preset variables can be clearly divided into two categories, resulting in the adoption of csQCA analysis. The basis of csQCA is the Boolean algebra. The main conventions are as follows: an uppercase letter represents the 1 value, read as “the variable is large, present, or high…,” while a lowercase or 0 value is the opposite. Logical “AND” is represented by the “*” symbol, and logical “OR” is represented by the “+” symbol. Conditions preceded by a “~” symbol are negative (equal to 0), while the ones that do not are positive (equal to 1). The symbol “→” is used to express the link between conditions and the outcome.
QCA is generally divided into three steps: data table construction, truth table reduction, and logical reduction (Denford et al., 2015). In the first step, all conditions must be calibrated as either “1” or “0.”Table 2 presents the calibration criteria. The second step is constructing a “truth table,” which is a table of configurations. Configuration is a given combination of conditions associated with a given outcome (Rihoux & Ragin, 2008). The third step yields three solutions—parsimonious, complex, and intermediate. Causal configurations arising in one of these three solutions may differ from those in another, but these configurations are always equal in terms of logical truth and never contain contradictory information (Roig-Tierno et al., 2015). The fsQAC3.0 software tool was used for the step-by-step operations.
Overview of Calibration (QCA).
The design of the conditions was mainly based on previous studies related to rumor communication using QCA methods (H. R. Wu & Luo, 2019; F. Zhang & Zhang, 2017). As the publisher and channel are not unique, this article considers the original and earliest publisher and channel. The outcome is “Effect,” which means the netizens’ attention to the rumor. Based on the search volume for keywords by netizens, most studies use the Baidu Search Index to reflect the effect. The Search index is based on the search volume of netizens in Baidu, using keywords as the statistical object, and scientifically analyzes and calculates the weighted sum of the search frequency of keywords. Specifically, they use the average value of 1,000 as a criterion for classification—above which is high impact and below which is low impact (J. S. Zhang & Qi, 2020; Zhou & Wang, 2016). This paper applies this standard. Some rumors do not have clear dates. Therefore, this paper uses the average daily search volume for the year in which the rumor spread. The description and assignment of eight QCA variables for scientific rumors are presented in Table 2.
The data for this study were primarily coded using one coder. To test the reliability of the coding, two other coders recorded 40 rumor samples selected using random sampling. We use SPSS software. The Fleiss’ kappa inter-rater reliability value of the Level 1 variables ranged from 0.687 to 1, with a mean of 0.833 (Supplemental Table S2), indicating good agreement.
Results
Overall Description of Scientific Rumors in China
Supplemental Table S3 presents the results of the content analysis, which analyzed a sample of 206 rumors. The themes of the rumors are concentrated in medicine and health (49.1%, 101 cases), food safety (27.7%, 57 cases), progress in science and technology (10.2%, 21 cases), and the environment and mysterious phenomena (10.6%, 22 cases). One marked feature of the theme of the rumors is that 76.8% (158 cases) of the samples are about health and safety. Of the rumors, 37.4% (77 cases) contain all six narrative elements, 40.7% (84 cases) contain four or five, while only 21.8% (45 cases) have less than four. Most of these rumors use a visual format (61.2%, 126 cases), including pictures and videos. These media make the presentation vivid and intuitive. Only 38.8% (80 cases) are expressed solely in text.
Rumors from unidentified netizen claims or homemade experiments are the most common (35.9%, 74 cases); these often include detailed real-life experiences of individuals or people close to them. Moreover, many rumors are drawn from highly credible sources, such as expert assertions (20.9%, 43 cases) or scientific research results (19.4%, 40 cases). However, the problems are the inclusion of vague descriptions, out-of-context quotations, and conceptual misunderstandings. The top three channels from our sample include WeChat (39.8%, 82 cases), websites (32%, 66 cases), and blogs (13.6%, 28 cases). The proportions of rumors originally from TV media were lowest—only 5.8% (12 cases). Regardless of the platform of the initial release, these rumors were fully spread online later. Social media platforms accounted for 62.2% (128 cases) of rumor sources. Specifically, some rumors first appeared in news reports and were then widely spread through WeChat. Examples include “pork has a hookworm that cannot be cooked or killed” and “spraying perfume while smoking can cause a car to explode on the spot.”
We found that 6.3% of all the rumors (13 rumors) persisted for more than two consecutive years (Table 3). Most of these were related to health and safety, such as cancer, smog, poisoning, radiation, infertility, and other topics closely related to people’s daily health and lives. This persistence reflects the resilience and impact of such rumors.
Recurring Scientific rumors.
Conditions of High-Impact Scientific Rumors
According to annual statistics from the China Internet Network Information Center (http://www.cnnic.net.cn/hlwfzyj/hlwxzbg/), since 2016, more than 95% of all China’s Internet users have been mobile Internet users. The years before 2016 reflected a period of rapid growth in the number of mobile Internet users. Therefore, we selected 2016 to 2020 for QCA (137 rumors included) to minimize the impact of changes in the size of Internet users on the number of rumor searches in the Baidu Search Index.
Necessity Analysis
In QCA, consistency and coverage are important indicators used to determine the degree of association between conditional and outcome variables. Consistency assesses the degree to which instances of an outcome agree in displaying the causal condition thought to be necessary, whereas coverage assesses the “relevance” of the causal condition—the degree to which a cause or causal combination “accounts for” instances of an outcome (Ragin, 2006). For example, when the coverage rate is 0.9, the variable can explain 90% of the cases. When the consistency score is more than 0.8, the condition variable is considered to be a sufficient condition for the outcome, and >0.9 is a necessary condition (Ragin, 2008; Schneider & Wagemann, 2012).
Before analyzing condition sets, we must first determine the “necessity” of each. As presented in Table 4, the consistency of “Detail” exceeds 0.9, indicating that it is a necessary condition for determining high-impact scientific rumors. Among the variables, “Detail” has the strongest effect. The consistency of “Fear” is close to 0.8, indicating a strong impact.
Analysis of Necessary Conditions.
Condition Combination Analysis
Figure 1 depicts the csQCA analysis of intermediate solutions, which cover a wider range, including logically possible combinations. Each row represents a combination of causal conditions. After excluding configurations with consistency below 0.9 and low coverage, the top five configurations with higher coverage were selected.
csQCA output.
Table 5 provides a clearer visual representation of Figure 1. When presenting notations, most authors use black circles (●) to indicate the presence of conditions and circles with “
” (⊗) to indicate its absence. Blank space indicates no need to care about this condition. Blank cells represent ambiguous conditions (N. Pappas & Glyptou, 2021; Ruiz-Mafe et al., 2018). Some scholars also use other notations, such as squares “□” and circles “○” (Roig-Tierno et al., 2015). In this paper, instead of using another notation, we simply use 1 or 0 to correspond to the conditions in Table 2. The sequence of each configuration is rearranged according to the level of raw coverage.
Crisp-set QCA Output.
Note. The raw coverage represents the proportion of cases that can be explained by combining the total cases. The unique coverage represents the proportion of cases that can only be explained by combining the total cases.
Configuration 1 has the highest coverage, covering the largest sample of rumors, and is considered the core configuration. It indicates that detailed rumors that appeal to fear, even if they are not published by the official media, have no reliable source, and spread through weak ties channels, can achieve a high impact. This highlights the role of detailed narrative elements and appeals to fear. Moreover, three-fifths of the configurations address this combination, coupled with the high consistency of each of the two conditions. This indicates that this specific combination has the strongest explanatory power for high-impact scientific rumors.
Configuration 2 is a combination of the following conditions: private media, visual presentation, unreliable source, detailed information, and resorting to fear, resulting in the high impact of rumors. Compared with configuration 1, this configuration adds a visual presentation. It is worth noting that four-fifths of the configurations involve this condition, and these configurations include 73.3% of rumors. This variable also has a consistency of 0.6395, which is close to 0.8, illustrating its strong effect on high-impact rumors.
Configuration 3 is the only one involving state media, while the other four configurations emphasize private media. Furthermore, the consistency of state media is the lowest in the analysis of necessary conditions, indicating a limited effect on high-impact rumors. Configuration 3 indicates that when a scientific rumor without special terminology is published by state media and shared in a weak-tie platform, combined with a visual presentation, authoritative sources, and detailed narrative elements, it can become a high-impact rumor.
Configuration 4 emphasizes the strong-tie channel, accompanied by a visual presentation and detailed narrative elements. Configuration 5 is the only one that includes scientific terminology, and its coverage is low, suggesting that the presence of scientific terminology is less conducive to the spread of rumors.
Conclusions and Discussion
The first part of this study provides a basic description of online scientific rumors in China using content analysis. The results reveal that the scientific rumors mainly concentrated on health and safety topics in China. Moreover, the largest number of rumors are those from experts, with 50.3% of rumors containing scientific terminology. Why? It was found that people tried to use scientific arguments and certified experts to gain credibility (Venkatraman et al., 2015). We find that 61.2% of these rumors use a visual format and 78.2% contain 4 to 6 narrative elements. Rumors were mostly text-based in the past but now appear in a visual format that users believe in more readily and share with others (Sundar et al., 2021). In addition, 6.3% of rumors persisted for more than two consecutive years. It has been found that when people are repeatedly exposed to the same rumor, they tend to believe it regardless of its truth due to the mere exposure effect (Lee & Kim, 2022). People believe repeated information more than novel information—a phenomenon called the repetition-induced truth effect (Unkelbach et al., 2019). Moreover, Shin et al. (2018) found that most false rumors were repeated periodically, while true rumors were not. Rumor management should be alert to the repetition of rumors.
Within our sample, private media was found to be the most frequent publisher of popular rumors (69.4%), especially social media (62.2%). Some scholars have also found that social media is a major source of rumor (Banakar et al., 2021; R. Wang et al., 2020). However, 30.6% of the rumors came from the state media. Guo (2020) found that official news websites advance the spread of rumors in China because of the weak enforcement of media regulations and the commercialized aspect of the media industry. First, the focus of regulation in China is on political rumors, and local media is more loosely regulated than central media. In China, in the past, only the government could create media. All traditional media are state media, both at the central and local levels. Second, since the introduction of the web, many experts have forecast the end of old media (such as television, printed books, and newspapers) (Ballatore & Natale, 2016). Peters et al. (2014) noted that one of the reasons for the economic crisis of traditional media is the loss of their quasi-monopoly over providing time-sensitive, topical information to a broad public. Therefore, some state media may fail to carefully vet information because they want to be faster than online media. Third, Wilson (2000) noted that accuracy in reporting and journalist ignorance were strongly related. Most Chinese science journalists, especially in state media, have majored in arts rather than science (Fu, 2019; Z. Wang & Wu, 2011); this may lead to a lack of scientific knowledge and literacy, which may also lead to the spread of rumors.
The second part of this study explores the factors that influence high-attention rumors (e.g., those receiving the most attention from netizens) using csQCA analysis. It revealed that a combination of detail and fearmongering are significant influencing factors associated with rumors that receive high attention. Specifically, detail is a necessary condition for high-impact rumors. Seah and Weimann (2020) also found that the details of a rumor are important incentives for audience forwarding. To enhance the content’s credibility, rumors usually adopt a narrative approach, which provides details of specific foods, people, time, place, and cause (Seah & Weimann, 2020). However, to some extent, the finding diverges from the research on the basic law of rumors, which highlights that ambiguity caused by the sketchiness of news is directly proportional to the intensity of a rumor (Allport & Postman, 1947). In the past, the limited memory capacity of our brain meant that rumors needed to be simplified when passed by word of mouth (Buckner, 1965). However, today, the Internet has minimized the need to rely on memory. For example, the “copy” and “paste” function in electronic tools makes it simple to keep rumors intact. Chen (2009) argued that the forgettable nature of oral communication allows vague rumors to be accepted, whereas today, overly vague internet rumors are usually not recognized. In general, the availability of more information is associated with greater effectiveness in reducing uncertainty and increasing the intention to trust and share (A. Y. K. Chua & Banerjee, 2016). Moreover, the accessibility and abundance of information have increased significantly on the Internet, facilitating the verification and refutation of rumors. As such, sketchy information may raise suspicions and checks.
Regarding fear, 70.4% of the scientific rumors in the study include fear tactics, and the QCA analysis indicates that the variable is highly relevant to high-impact rumors. Buhl et al. (2019) found that importance, deviance, and proximity made no difference in the initial diffusion processes of rumors, while damage was positively associated with fast diffusion. The samples in this article often promote fear by emphasizing damage to the body. Specifically, risk words such as “radiation,”“toxic,” and “dead” appear frequently. In addition, many rumors mention diseases or epidemics that cause extreme pain or death, including cancer, leukemia, plague, and dengue. To some extent, user engagement, retweets, and comments increase significantly as the level of fear-arousing sensationalism rises (Ali et al., 2019). For example, one study found that the first and most viewed Google result images related to GMOs contain the most frightful content (Ventura et al., 2017). Regarding the high relevance of fear, a possible factor is that humans are inherently interested in information about environmental life threats (Shoemaker, 1996). Moreover, a meta-analysis by Witte and Allen (2000) suggested that a strong fear-based appeal produces high levels of susceptibility, increases the perception of severity, and is more persuasive than low or weak fear appeals. Moreover, newsworthiness increases if identifiable events can be linked to a threat to human life. To do this, levels of alarm are often magnified in media coverage (Weingart et al., 2000). The media’s emphasis on fear tactics may influence people’s recognition and behavior.
Furthermore, this study reveals that the presence of scientific terminology is less conducive to the spread of rumors. A minority of people believe that they are more likely to be deceived by rumors if the rumors need to be clarified by professional scientific knowledge (Zeng et al., 2021). “Receiver Costs” are an important component in communication dynamics: if a reliable signal is very costly to assess, receivers may choose to rely on one that is less reliable but easier to obtain (Donath, 2007; Guilford & Dawkins, 1991). Science has become more difficult for non-specialists to understand (Hayes, 1992). However, in 2020; only 10.56% of Chinese were classified as scientifically literate (L. He et al., 2021). Thus, a failure to understand scientific terminology may be the reason for its limited impact.
Our QCA analysis also reveals that state media has a weak effect on high-impact rumors. First, this may be related to the public’s distrust of the government. Lee and Kim (2022) found that government trust moderated the relationship between anxiety and rumor exposure in both rumor cases. In China, health and safety were the most common themes in our rumor samples. However, many food safety incidents or scandals in China (e.g., the melamine milk scandal and the “gutter oil” scandals) have significantly undermined consumer trust in the government (Kendall et al., 2019; R. Wang et al., 2020). Those with low levels of trust tend to prefer non-mainstream news sources (Fletcher & Park, 2017). Kendall et al. (2019) found that due to the lack of trust in the regulatory environment, Chinese customers rely on informal kinship networks as trusted sources.
Second, people may not care or be able to distinguish between state and private media. Individuals often assume that statements they believe originate from credible sources (Fragale & Heath, 2004). In addition, in China, only the government could create media in the past; now, in sharp contrast, everyone can publish and communicate information using different kinds of online media. The many forms and mixed nature of media may confuse people.
Third, the standard of determining credibility has changed. Lankes (2007) proposed that models of credibility have shifted from traditional authorities to what is called a “reliability approach,” where the user determines credibility by synthesizing multiple sources of credibility judgments. Traditional notions of credibility are defined as coming from a centralized authority (e.g., government or expert). Electronic networks make it easier to rely on the collective to assess information (Flanagin & Metzger, 2008). As such, this may diminish the credibility of the state media.
We identified a few limitations of this study. First, as the distribution of rumors in this article spans a long period, the search volume on the outcome variable may be influenced by some potential factors, such as the number of netizens in different years. Second, scientific rumors may be characterized by more than the seven variables described in this paper. More unique characteristics of scientific rumors can be identified in the future by referring to different theories. Third, we used the selected ranking list to identify rumors that are more likely to be spread on the Internet, without investigating users’ perspective. As such, the analysis of the influencing factors was not based on user attitudes and behaviors. Finally, instead of investigating the audience’s feelings about rumors, we hypothesized that the factors were based on their wide circulation. To summarize, we emphasized the content characteristics and the factors influencing scientific rumor communication. More attention should be paid to rumors that provide detailed and vivid information, appeal to fear, and are spread through personal social networks. Nowadays, everyday consumers of media are no longer passive but active recipients of this content; therefore, future research can be conducted from the perspective of the audiences who read, believe, and spread scientific rumors.
Supplemental Material
sj-docx-1-sgo-10.1177_21582440231215586 – Supplemental material for Too Real to be Questioned
Supplemental material, sj-docx-1-sgo-10.1177_21582440231215586 for Too Real to be Questioned by Lingfei Wang, Mengmeng Yue and Guoyan Wang in SAGE Open
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research was supported by the National Natural Science Foundation of China (grants 82273744).
Ethics Approval and Consent to Participate
Not applicable.
Consent for Publication
Not applicable.
Supplemental Material
Supplemental material for this article is available online.
Data Availability Statement
Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
