2.1. Social Media as an Information Source

Social media’s utility as an information source is ever-increasing. Shearer and Gottfried (2017) found that 74% and 68% of users get information from Twitter and Facebook. According to Broersma and Graham (2012), Twitter has an impact on political campaigns and elections. Many Facebook users consider it as a valuable source of information (Lampe, Vitak, Gray, & Ellison, 2012). Bene (2017), for example, found that many young people use Facebook as a primary source of political information. The study also hints that opinion leaders’ roles are important in this information-seeking process. While many national media outlets were enjoying shut downs across the Arab world in 2011, Facebook and Twitter, being the primary means of communication and networking, brought a whole new revolution: We call it the Arab Spring (Eltantawy & Wiest, 2011; Khondker, 2011; Wolfsfeld, Segev, & Sheafer, 2013). Apart from the political domain, social media information is used for healthcare and medical purposes as well. Social media campaigns, for example, were effective in educating US citizens about the dangers of the Zika virus (Southwell, Dolina, Jimenez-Magdaleno, Squiers, & Kelly, 2016). Similar social media campaigns are also seen amid the COVID-19 pandemic. However, rumor propagation in social media amid such pandemics may hinder the campaigns’ positive outcomes and provoke mass anxiety. Social media is equally important for medical education. Medical students, for example, produce and share information with their peers in social media for study purposes (Zulfikar, Zaheer, Baloch, & Ahmed, 2018). Social media users often exchange emotional information instead of factual information, which helps them to create a collective emotional state. Such emotional expressiveness is also the fuel that helps some e-communities to sustain themselves (Chmiel et al., 2011). Further, some users produce provocative and emotional information to attract other users. The lack of gatekeeping often lets users disseminate extreme ideologies to misguide netizens, such as religious fanaticism which could lead to interreligious disharmony (Abouzakhar, 2015; Al-Zaman, 2020b).

2.2. Rumors and Their Typologies

Rumors can be both an initiator and an outcome of information pollution. Meel and Vishwakarma (2020) present a taxonomy in their study in which they mention about ten categories of false information: rumors, fake news, misinformation, disinformation, clickbait, hoaxes, satire/parody, opinion spam, propaganda, and conspiracy theories. However, they also state that a rumor is only unverified information that may not be false in some cases. In contrast, Seo, Mohapatra, and Abdelzaher (2012) think a rumor means false information, and true information is a non-rumor. In general, a rumor is “a proposition of belief, is officially unverified when issued, and should deal with either current events or topical issues to express the emotional needs of the community and/or to help people make sense in the context of ambiguity, danger or potential threat” (Kim, Bock, Sabherwal, & Kim, 2019, p. 593). Such commonly arises not only in ambiguous and threatening situations but also when information is scarce and people feel a psychological need for understanding or security (DiFonzo & Bordia, 2007b; Watson & Hill, 2006). When uncertainty and ambiguity are reduced to a certain degree and reliable information is provided, rumors disappear. Thanks to their prominence, rumors are often more difficult to disprove than to prove (Dowd & Davidhizar, 1997). A rumor’s strength varies with a story’s thematic importance and topical ambiguity (Allport & Postman, 1946). Shibutani (1966) adds that some mental/emotional arousal is required for rumors’ production, circulation, and success. This is why a rumor is called an infection of the mind (Nekovee, Moreno, Bianconi, & Marsili, 2007).

Several studies present typologies of rumors from different perspectives. For example, Chua, Aricat, and Goh (2017) classify rumors based on three criteria: the level of its direness, its relationship with unquestionable truths, and its passionate request. They analyze the life-cycle of rumors and find that rumor messaging is mainly of four types: information-related, emotion-related, deliberation, and call-to-action (p. 263). Rumors can also take the form of a hoax, joke, little story, or information leak (Lewandowsky, Ecker, Seifert, Schwarz, & Cook, 2012; Rosnow, 1991). It can be breaking news as well as a form that lacks sufficient proof or supporting evidence (Turenne, 2018). Kapferer (2013) proposes a typology that includes six types of rumors, based on five cross-sectional factors. For him, two basic categories of rumors are spontaneous and provoked. According to Dowd and Davidhizar (1997), a rumor has four types: wish, fear or bogey, wedge-driving or aggressive, and anticipatory. However, Chua et al. (2017) exclude the anticipatory rumor from their modified typology. Wish rumors look for prospects. They can attract the attention of the audience with their irrational and fanciful representation of future benefits and hopes (Chua et al., 2017). Dread rumors usually stir up fear (Rosnow, 1991; Rosnow, Esposito, & Gibney, 1988). Such rumors are relatively higher in number, have longer life spans, and have higher persuasive value compared to wish rumors (Walker & Blaine, 1991). Wedge-driving rumors tend to polarize societies, fueled by stereotypes and hatred (Bhavnani et al., 2009; Sunstein, 2014).

Derczynski et al. (2015) provide a more comprehensive typology of rumors. According to their study, a rumor has four categories: speculation, controversy, misinformation, and disinformation. Speculation is simply the consumption of an asset with the hope that it will become more valuable shortly. For example, despite knowing the risk, someone buys stock market shares. Controversy is a state of public dispute and a matter of conflicting opinions (Derczynski et al., 2015). Misinformation is typically unintentional and inaccurate information mainly arises from knowledge gaps. It includes false datasets, insults, and pranks. Disinformation is also known as false information, but unlike misinformation, it is produced and disseminated deliberately to deceive people (Al-Zaman, 2019). Often, the four terms: rumor, misinformation, disinformation, and fake news, are used interchangeably as they share some common areas (e.g., Duffy, Tandoc, & Ling, 2019; Tandoc, Lim, & Ling, 2020; Wardle & Derakhshan, 2017. For example, as some research defines fake news as disinformation (Tandoc et al., 2020), some other studies address it as misinformation (as cited in Tandoc, Lim, & Ling, 2018). This means fake news can be both intentional or unintentional false news/information. Also, fake news and rumors are considered two closely-related components in the information ecology (Duffy et al., 2019). In this study, we prefer to use the term rumor to avoid conceptual discrepancies.

Apart from the discussed typologies, little research has been conducted so far to understand the theme- or topic-based categorization of rumors. Based on the critical-historical lens of media ecology analysis, Higdon (2020) explores four themes of rumors: nationalism, hate, celebrity gossip, and fear. Bordia, Jones, Gallois, Callan, and DiFonzo (2006) present six themes by analyzing change-related healthcare rumors. They include changes to job and working conditions, the nature of organizational change, poor change management, consequences of the change for organizational performance, gossip-rumors, and un-coded statements. In another study, Wu and Liu (2018) identify four types of rumors in social media: business, science and technology, entertainment, and medical. This study limits its extent within a more technical genre, i.e., computer science, and it also does not explain the categories.

2.3. Sources of Rumors

Apart from typology, sources are equally important in the study of rumors. Source identification is common in media and Internet research, such as finding the source of a virus (Shah & Zaman, 2010), sources in social and communication networks (Comin & da Fontoura Costa, 2011), sources in mixed networks (Jiang, Wen, Yu, Xiang, & Zhou, 2017), and sources of misinformation in social networks (Nguyen, Nguyen, & Thai, 2012). However, source identification of social media rumors gets less attention from scholars to date. Kapferer (1992) in his micro-level analysis discusses eight primary sources of rumors from where they were born. These include experts’ opinions, confidential information, troubling facts, testimony, fantasies, urban legends, misunderstandings, and manipulation. On the other hand, Jo (2002) categorizes the sources of Internet rumors into two main types: the Internet itself and traditional mass media. The Internet as a form of media, according to the study, includes different online information sources such as websites and social media, and traditional media includes television, newspapers, and radio. The study also shows that rumors from online sources increase with time, while rumors from traditional media decrease. In a more relevant study, Shin, Jian, Driscoll, and Bar (2018) explore the two main sources of rumors by analyzing their Twitter dataset. As most of the rumors on Twitter are from non-traditional media, a few are from traditional media. They define non-traditional media as websites, social media, and other Internet sources. Similarly, Muigai (2019) finds two sources of social media rumors: online media and mainstream media. The study further suggests that most of the social media rumors originate from online media.

2.4. Aims of Rumors

Several studies analyze the aims of rumors. Although the words “aim” or “intention” may sound problematic, a few empirical and experimental studies use these terms to describe the tendencies and claims of rumors. For example, Kamins, Folkes, and Perner (1997), surveying consumers to understand marketplace rumors, find two types of aims: positive and negative. In the survey, the consumers reported that they are exposed to and transmit more negative rumors than positive rumors. The survey findings were tested in two further experiments to validate the results. Knapp (1944), on the other hand, also finds positive and negative aims of rumors. According to him, a positive rumor usually predicts wishful or positive outcomes, while a negative rumor describes undesirable or harmful outcomes. Bordia, Jones, Gallois, Callan, and DiFonzo (2006) also find similar aims in their study on health-related rumors. They show that negative rumors (n=479; 93.92%) are more prevalent than positive rumors (n=31; 6.08%) in their sample. Another study shows almost similar results, i.e., negative rumors (91%) are more commonplace in social media than positive rumors (9%) (Avaaz, 2019). Rosnow, Yost, and Esposito (1986) also classify rumors as positive and negative but based on their consequences. Some more studies show almost similar findings that are based on empirical data and experiments (for more, see Ji, Yan, & Yu, 2020).

4.1. Categories of Social Media Rumors

The analysis of rumors produces seven theme-based categories (see Table 2). Political (n=62; 34.3%) is the most prominent category, followed by health & education (n=29; 16.0%), crime & human rights (n=20; 11.0%), and religious (n=20; 11.0%) categories. Political rumors are mainly related to elections, political parties, national and international political issues, important political events, development initiatives, and political figures. Health & education rumors are mainly related to the healthcare system, accidents and deaths, diseases, primary and higher education, and academic corruption and dishonesty. Homicides, rapes, abductions, ethnic cleansing, and violence are the important focuses of crime-related rumors. Religious rumors include mainly religious conversations, religion’s significance, religious laws and regulations, and historical religious incidents. It should be noted that religious and religiopolitical rumors together comprise 20.99% (n=38) of the total share, which could be the second-highest after political rumors. This suggests that religion is a convenient topic for rumor propagation in Bangladesh.

Table 2.

Categories of rumors

Of seven categories, political rumors seem more dominant than others throughout the timeline. In 2017, political rumors was the highest, which was 43.5% of the total rumors (see Table 3). In 2018 and 2019, its share was 56.2% and 32.0%, respectively. Political rumors within its category reached its peak (66.1%) in 2018. The surge of political rumors in that year was due to the 11th Parliamentary Elections. However, the share of political rumors dropped to only 5.0% in 2020; the percentage decreases every consecutive year. More interestingly, health-related rumors surged from 0% in 2019 to 41.7% in 2020, and most of the rumors are related to the COVID-19 pandemic, such as “Hot weather prevents the spread of coronavirus” and “Vaccine is made for coronavirus.” In the same year, health & education, religion, and crime & human rights rumors reached their highest numbers within their categories, which are 86.2%, 40.0%, and 55.0%, respectively. On the other hand, religiopolitical, entertainment, and other rumors reached their peaks in 2018 with 50.0%, 47.4%, and 30.8%, respectively. Religious rumors are increasing in every consecutive year, from 5.0% in 2017 to 40.0% in 2020, while others fluctuate.

Table 3.

Different categories of rumors in 2017-2020

4.2. Sources of Social Media Rumors

Social media rumors have two main sources: mainstream media and online media (see Table 4). The mainstream sources, including television channels, newspapers, and radio channels, are mostly national media outlets. On the other hand, online sources include online versions of mainstream television channels and newspapers, online news portals, blogs and websites, and social media. Most of the social media rumors we found in our sample emerge from online media (n=148; 81.8%) and a smaller amount from mainstream media (n=33; 18.2%). In 2017, online media was responsible for 65.2% of social media rumors, whereas mainstream media was responsible for 34.8% of rumors (see Table 5). The gap widens in 2020 with 93.3% online media-based rumors and only 6.7% mainstream media-based rumors. However, this gap was not widening constantly, rather it fluctuated. In 2018, for example, rumors from online and mainstream media were 80.8% and 19.2%, respectively. In 2019, online rumors decreased to 72.0% and mainstream media increased to 28.0%. The percentages changed again and moved in the opposite direction in the next year. Also, the percentages of online rumors ebbed and flowed in every consecutive year, which are 10.1%, 39.9%, 12.2%, and 37.8% in 2017, 2018, 2019, and 2020, respectively. Although the numbers fluctuate, the last two years show an upward tendency. Conversely, the percentages of mainstream media rumors decrease with a slight fluctuation: 24.2%, 42.4%, 21.2%, and 12.1% in 2017, 2018, 2019, and 2020, respectively. Unlike online media, the last three years of mainstream media show a downward tendency. Both mainstream (27.3%) and online media sources (35.8%) produce political rumors in the highest numbers (see Table 6). From mainstream and online media, the least produced rumors are crime & human rights (3.0%) and entertainment (4.7%) rumors. The sources of entertainment rumors seem balanced with 46.2% of mainstream media and 53.8% of online media, whereas mainstream (5.0%) and online (95.0%) media have a huge gap in terms of crime & human rights rumors. Most interestingly, of online media most of the rumors (n=94; 63.51% of online sources and 51.94% of all sources) are produced in three social media platforms: Facebook (n=83), YouTube (n=7), and Twitter (n=4).

Table 4.

Sources of rumors

Table 5.

Sources of rumors in 2017-2020

Table 6.

Sources of different categories of rumors

4.3. Aims of Social Media Rumors

The dataset contains n=132 negative rumors, which is 72.9% of the total cases (see Table 7). In contrast, positive rumors are only n=41, which is 22.7% of the total. The leftover n=8 (4.4%) rumors show seemingly no inclination; thus they are included in the unknown category. Among the rumor categories, health & education-related rumors (86.2%) are mostly negative over other counterparts: political 75.8%, religious 75.0%, religiopolitical 72.2%, crime & human rights 70.0%, other 57.9%, and entertainment 53.8% (see Table 8). Also, of negative rumors, political rumors are on the top of the list. Although a large share of political rumors (35.6%) are negative, the aims of the most political rumors (37.5%) cannot be identified. Negative aims are found less in entertainment rumors (5.3%) compared to others, but most of the entertainment rumors are negative (53.8%) rather than positive (38.5%). Positive aims are found more in political rumors (29.3%) and less in health & education rumors (9.8%) and religiopolitical rumors (9.8%). Also, online media is a more influential source for negative rumors (86.4%) over mainstream media (13.6%), where online rumors are almost sevenfold of mainstream rumors (see Table 9). This huge gap suggests online media produces a higher number of perilous rumors in contrast to mainstream sources. Online media produces more negative rumors (77.0%) than positive rumors (18.9%). In this case, mainstream media is comparatively better than online media as it produces 54.5% of negative rumors and 39.4% of positive rumors. Like the negative rumors, the gap between online media (68.3%) and mainstream media (31.7%) for positive rumors is relatively lower, i.e., online rumors are more than twice in number of mainstream rumors. Rumors with unknown aims have both online (75.0%) and mainstream sources (25.0%). Interestingly, the gap (that is 58.1) between negative (77.0%) and positive (18.9%) rumors from online media is more than the gap (that is 15.1) between negative (54.5%) and positive (39.4%) rumors from mainstream media. It further shows how intensively online media produces negative rumors as opposed to mainstream media.

Table 7.

Aims of rumors

Table 8.

Aims of the different categories of rumors

Table 9.

Aims of rumors from different sources

5.1. Main Objectives

The primary aim of this study was to analyze social media rumors. Three specific focuses were to identify the theme-based categories, tentative aims, and prime sources of social media rumors. First, N=181 social media rumors’ data between 2017 to 2020 in Bangladesh, a rumor-prone South Asian country, were collected. Second, the collected data were analyzed following the content analysis method. The results show that the seven theme-based categories of social media rumors are political, health & education, crime & human rights, religious, religiopolitical, entertainment, and other; the two main sources of social media rumors are mainstream media and online media; the three tentative aims of social media rumors are positive, negative, and unknown.

5.2. Key Findings

The study has a few key findings. First, of seven theme-based categories, political rumors are the highest-produced social media rumors, while entertainment rumors are the lowest-produced rumors. Production of political rumors in social media is influenced by a few factors, such as election and political events. It is more likely to surge before and during an election period. Studies such as Faris et al. (2017), Posetti and Matthews (2018), Silverman (2016), Parkinson (2016), and Serban et al. (2017) also suggest the production and distribution of social media rumors ahead of and during an election. In the absence of significant political events, political rumors follow a moderate trend. Interestingly, the percentages of political rumors in social media were found to be decreasing every consecutive year. This could be a consequence of the country’s increasing political stability through a hybrid political system (Riaz, 2019). Second, health-related crises stir up mass anxiety that further expedite the production and distribution of health-related rumors in social media. This may support the thesis of Allport and Postman (1946), Difonzo and Bordia (2007a), and Watson and Hill (2006), i.e., rumors are produced amid ambiguity, uncertainty, scarcity of information, and psychological insecurity. With the prevalence of the coronavirus, recent rumors are mostly pandemic-related that either tend to satisfy the emotional needs of the community or make sense of health-related issues, according to Kim et al. (2019). However, in many countries like Bangladesh, health and education rumors may not be so commonplace. Third, religion-related rumors are increasing in social media. In Bangladesh, rumors related to Islam, the dominant religion in the country, are increasing. Religious rumors are not so harmful like religiopolitical rumors, but they may work as an ideology- and identity-shaper.

Fourth, of the two sources of social media rumors, online media produces more rumors than mainstream media does. More specifically, online media is the prime source of social media rumors, which supports the findings of Jo (2002), Muigai (2019), and Shin et al. (2018). Moreover, social media itself contributes to rumor production to a large extent. Several studies also deal with social media’s rumor production and circulation capacity (Goh et al., 2017; Hashimoto et al., 2011; Kwon et al., 2013a, 2013b; Oh et al., 2010). However, these studies either do not specify social media’s contribution in rumor production, or do not present a comparative analysis of mainstream and online sources of rumors, or do not position social media rumors in the broader context of online media. Fifth, online media produces a lower amount of entertainment rumors, while mainstream media produce a lower amount of crime rumors. However, both sources have two similarities. One, online and mainstream media tend to produce a higher number of political rumors than other types of rumors. Two, the percentages of online and mainstream rumors fluctuate over the years, making it difficult to predict their future trend.

Sixth, most of the social media rumors have negative aims that outnumber positive aims. This finding is consistent with the finding of Avaaz (2019) and Bordia et al. (2006). However, the second study is health-related and does not take social media into account. It may be inferred, therefore, that the proportions of aims remain almost constant irrespective of media or context. Both positive and negative aims are higher in political rumors than other types of rumors. However, negative aims are more evident in political rumors than positive aims, which might be alarming for the political environment. Similarly, health-related rumors are negative in most cases. As mentioned earlier, the COVID-19 pandemic has a serious contribution in the burgeoning health-related negative rumors that emerge from information scarcity, situational uncertainty, and mass anxiety, mentioned in detail in the studies of Allport and Postman (1946), Difonzo and Bordia (2007a), and Watson and Hill (2006). Meanwhile, a single COVID-19-related rumor claimed at least 800 lives worldwide from January to March 2020 (Coleman, 2020).

5.3. Limitations and Implications

This study has a few strengths and limitations. First, the time durations for data collection were not equal for all the years (see Table 1). The only reason is that this study was conducted in mid-2020, which makes it virtually impossible to take the whole year into account. Second, crosstabulations may not effectively describe the interrelationships between the variables, but it helps to predict the trend to some extent. We suppose that correlation coefficient analyses among the variables would provide some better insights. Third, findings from many relevant studies were not possible to incorporate into this research because of time and resource constraints. Fourth, the data for this study was collected from a fact-checking website that is seemingly a reliable source for rumor data. However, two issues are important to note: Fact-checkers often fail to address all the rumors due to their limited resources and accesses (Brennen et al., 2020), and data from a single source may not present a holistic scenario. Therefore, for a more efficient future study, more extensive data from various sources would be required. Fifth, ample researches have been done so far that deal with rumor typologies but not from the perspective of the popular themes of social rumors. Also, source and aim analyses are scant in previous literature. In this regard, this study produces some novel findings and opens a new threshold for further researches in this area, offering a methodological and conceptual framework for online rumor analysis. The results may help to understand what types of rumor topics are more popular among netizens, produced from which sources with what aims, and to what extent they vary in different situations. More importantly, surging religion-related rumors suggest taking necessary precautions to restrict digital communalism (Al-Zaman, 2019). Similarly, proper information flow is essential during a crisis period (e.g., a pandemic) to restrain negative health-related rumors. The results may also guide relevant micro- and macro-level policymaking and strategic decisions in the context of Bangladesh. We also think, however, that a few research endeavors are still required to bridge some existing knowledge gaps. The South Asian region, as mentioned in the introduction, has become a hotspot of social media rumors, having detrimental impacts on society. Therefore, we would like to invite more research that would deal with who produces rumors, how they are produced, what are the popular distribution channels, who are the most desired target audiences, how people engage with rumors, what are their most frequent outcomes, and what measures should be taken to prevent this problem.