Keywords

1 Introduction

1.1 Current Theories in Literature

Since 2009, Social Network Sites (SNS) have gained attention not because of their function of creating and exchanging user-generated content but also because they have emerged as a main channel for information-seeking and news distribution. One indicator of this shift may be seen in Pew Research Center data: In 2018 68% of U.S. adults got their news via social media [1], whereas in 2012 only 49% did so [2]. Complementing the rise of SNS is the ever-increasing oceans of information available online, which can become a special problem if some of this information is misleading, intentionally wrong or falsely promotional [3]. Research on natural disasters, for instance, has shown that while on average 30% of total tweets posted about an event contained situational information about the event [4]. This percentage is even higher in health-related issues: of all anorexia-related videos posted in YouTube, 30% misinformed viewers about several aspects of this eating disorder [5]. In coverage of the HPV vaccine, one-third of the videos did not present accurate information [6]. Subsequent paragraphs, however, are indented.

Scholars who investigate the internet’s role in promoting misperceptions have offered a variety of explanations for user’s finding and sharing false information and news. One views is that online news facilitates politically biased news consumption [7], leaving partisan audiences with a deficit in their knowledge. This promotes inaccurate beliefs and the formation of “echo chambers” or “filter bubbles” [8]. Another view holds that misperceptions are the result of the psychological processes through which information is interpreted and then perceptions formed [9,10,11]. Additionally, researchers have looked at the phenomena as being influenced by political factors (e.g. strength of partisanship); derived from technical characteristics (e.g.. algorithms or social media use) and psychological-attitudinal (e.g., trust in media or conspiracy mentality) [12].

However, to the best of our knowledge, there are no comprehensive efforts to integrate these different dimensions into a single model in order to understand the relative weight and relevance of these factors. Accordingly, our study aims to assess which types of variables influence users to be exposed to, believe in, and share the misinformation they find online. We start our analysis by positing that in order to share fake news (i.e., that which the users interpret as “news” and not as an instance of irony or repudiation), users first need to believe in, or at least be open to, the general orientation represented by the false item and also be exposed to the specific item’s content. Thus, we suggest that social media qualities, political behavior/ideology, trust and psychological factors have an indirect effect on sharing fake news, mediated by exposure and credibility. We test our model through a Structural Equation Model (SEM), which integrates the above (operationalized) variables, on a two-wave Chilean panel data collected in 2017 and 2018.

1.2 Social Media Affordances

Used in the human-computer interaction, design, and communication fields [13], the term “affordance” refers to the perceived—and actual—properties of an object that allow people to do something with it. The literature suggests that SNS like Twitter or Facebook incorporate at least three set of affordances that facilitates the generation and spread of fake news through internet [10, 12, 14, 15]. First, platforms like Facebook or Twitter allow any users produce and share information at little or no cost. Second, given the vast amount of information produced via these platforms and the limited time users have to process the data they encounter, SNS are creating an environment where information exceeds the capability of users to reflexively analyze the news that they are exposed to. This is especially the case when they use their phones to interact with this content, further limiting their ability to evaluate the veracity of the shared information. Third, as several authors have argued, SNS are ideologically segregated by algorithms that filter information to which users are exposed based on their preferences and online behavior. Thus, saturated information spaces, low cost of information production, and algorithms, facilitate SNS’s role in circulating fake news.

Following this line of argument, we can hypothesize that those individuals who use SNS more frequently would also tend be more exposed to information through these platforms, and consequently, more exposed to fake news and, given the above factors, may be more likely to believe the fake news. On this basis, we argue:

  • H1: Frequency of social media use is positively related to (a) exposure to fake news, and (b) holding misperceptions (i.e., incorrectly believing false news are accurate).

  • H2: News consumption through social media is positively related to (a) exposure to fake news, and (b) holding misperceptions.

1.3 Political Participation and Identification

A second set of variables used to explain the proliferation of fake news is related to political behavior and ideology. Several studies have found a positive relationship between actively participating in politics and being exposed or believing in fake news [16, 17]. They hold that as users participate more actively in politics, they develop a political or party identity, which is linked to cognitive biases [18, 19]. Thus, political participation has been traditionally associated to the formation of closed groups that function as news filters that generate a selective exposure of information [20, 21].

Most of this literature draws on the cognitive theory developed by Kunda concerning confirmatory biases [22]. This perspective holds that individuals with marked political preferences tend to orient their conclusions to confirm prior political beliefs. More, Kunda found that while individuals seek to reach conclusions that coincide or reinforce their positions, they are not free to take any path to the conclusions they wish. Rather, that to reach certain conclusions, individuals must reasonably justify their positions, maintaining an “illusion of objectivity”.

Consequently, it may be more difficult for politically active individuals to distinguish the veracity of different media due to their cognitive biases that filters the type of news to which they are exposed. And, when it challenges their beliefs, cognitive bias also prevents them from considering corrective information in a balanced way.

Further, the relationship between political participation and sharing misinformation is affected by the level of misperception in individuals [23]. Researchers show that in explaining the individuals’ support in US for the Iraq war in 2003 for instance, in addition to party identification, it was also relevant what information they had about the possession or absence of weapons of mass destruction by the Iraqi government. Yet it was the mistakenly informed people who were most likely to share their opinion online. Drawing on their finding, it seems that, unlike people who are totally uninformed about a topic, misinformed individuals are more likely to share inaccurate information on SNS, especially when they are politically active. Given that this research seeks to compare the relevance of different types of variables, we include hypotheses of both participation and political ideology to evaluate the impact of political factors compared to other types of variables.

  • H3: Frequency of online political participation is positively related to (a) exposure to fake news, and (b) holding misperceptions.

  • H4: Extremity of political opinions is positively related to (a) exposure to fake news, and (b) holding misperceptions.

1.4 Trust in Media, Contacts and Conspiracy Mentality

The third set of possible explanations we consider assume that fake news proliferation through social media is a consequence of attitudes [24] and psychological factors [25], especially those associated with trust. Trust, which is understood for the purpose of this paper, is an element that allows people to overcome the vulnerability derived from uncertainty through competence, reliability, integrity and security of other people and systems [26]. The reason is simple: confidence works as an affiliative conduit that allows individuals to discern between sources of information, peoples or institutions, reliable or unreliable [14].

Yet confidence in institutions is declining in many areas of social life, such as traditional media, which can facilitates exposure to misinformation and its propagation [10, 14]. Three reasons have been advanced by researchers for this spread of misinformation/fake news. First, has to do with news media: Traditional news media are becoming indistinguishable from other forms of news dissemination, collectively called alternative news media, whose rigor is difficult to verify [11, 21, 26]. Following this logic, we propose the following hypotheses:

  • H5: Mistrust in traditional news media is positively related to (a) exposure to fake news, and (b) holding misperceptions.

A second area has to do with trust in social media contacts. These help define how users share and interact with content posted by other users [27]. If users trust their contacts, they (almost by definition) would trust the information shared by them. Further, studies have shown that the interactions that occur in platforms like Facebook work in a similar way to face-to-face interactions, as these interactions promote intimacy, confidence, and participation [28]. Consequently, it is expected that the more confidence users of social networks have in their contacts, the more credibility they will have in the information shared by them, including fake news. Therefore, given the increased circulation of fake news via SNS, it is expected:

  • H6: Trust in information shared by contacts is positively related to (a) exposure to fake news, and (b) holding misperceptions.

Third, psychological variables affect the proliferation of misinformation through social media, particularly the influence of conspiracy mentalities, which refer to claims that seek to explain some event or practice by reference to the machinations of powerful people, who attempt to conceal their role [29]. According to Flynn et al. [16] they are distinctive insofar as they focus on the behavior of powerful people and may be rooted in stable psychological predispositions. Others suggest that conspiracy mentality can be defined as a belief system related to a rejection of what is perceived as power groups who are covertly responsible for negative political or economic events [30]. Conspiracy theories have been a common topic among fake news studies [10, 14, 25]. Drawing on them, we expect that people who have a greater conspiracy-oriented mentality will be more likely to believe in rumors (or facts) unsupported by the standards of evidence, regardless their political ideology or partisanship. Taking this into account we propose the following hypothesis:

  • H7: Having a conspiracy mentality is positively associated with (a) exposure to fake news, and (b) holding misperceptions.

2 Methods

2.1 Data and Context

A two-wave panel survey was conducted to examine the hypotheses, during April 2017 and June 2018 respectively. A national panel was employed following the Chilean National Socioeconomic Characterization Survey (CASEN) in order to assure a more accurate national representation. Three variables were used to generate a representative sample: gender, age, and geography. Of the 8840 participants who received the initial email with the survey’s URL, 1007 respondents ended up participating in the first wave (2017) and of those 1007, 45% participated a year later in the second wave (451). To correct for demographic biases, we used a model-based strategy, which means that we entered as a control any variable that could be used to construct a post-stratification weight [31].

Concerning the country were the data was collected, it is relevant to mention two aspects. First, after Chile’s last political election (November 2017), the fake news issue became a subject of national media attention, which warned the public, inter alia, that fake news was seen 3.5 million times during the year 2017 [32]. Second, Chile has a high use of social networks such as Twitter, Facebook or WhatsApp, all of which serve as a source of information and news (64%) compared to the world average (23%) [33], and a growing distrust in traditional media such as television and newspapers [34]. Consequently, it is possible to argue that Chile also experiences the symptoms of informational disorder observed in the global North [35].

2.2 Analysis

We used a lagged dependent variable model estimated with OLS regression where exposure, credibility, and sharing fake news variables in wave two are the main outcome variables. This type of model is used under the assumption that the variable of interest, in this case exposure, credibility and sharing of fake news of wave 2, are strongly explained by their past (wave 1). Therefore, the exclusion of the lagged variables of wave 1 can lead to biases of omitted variable and reduce the reliability of the coefficients of the rest of the independent variables. The inclusion of lagged variables absorbs a large portion of the variance of the model, while the remaining coefficients can be interpreted as the predicted change in dependent variable compared with the value it could have taken knowing its previous value [36, 37].

We are aware that the inclusion of lagged dependent variables could reduce the contribution of other independent variables and increase standard errors. However, this quality makes our models conservative when estimating the coefficients, which gives more robustness to our conclusions.

We also use as a complement tools of structural equations (SEM) to identify direct and indirect effects of the independent variables in the three variables of interest separately. The data also presented a small proportion of missing values. So as not to bias the coefficients, we decided to impute these cases with the mean (N = 423).

Finally, all independent and dependent variables considered in the analysis except for lagged variables correspond to wave 2 of the survey. Figure 1 shows the theoretical model with proposed causal relations in which social media qualities, political behavior and ideology, and trust and psychological variables impact directly exposure and credibility to fake news. Of course, both variables then explain sharing fake news. So, in this article we do not focus in the direct relation between the first three groups of variables and sharing fake news, but in their indirect relation mediated by exposure to and credibility of fake news.

Fig. 1.
figure 1

The hypothesized variables included in the model

Full size image

2.3 Dependent Variables

Exposure to Fake News.

Respondents were exposed to a set of 14 fake news stories that circulated in Chile in the preceding 15 months and then were asked if they were aware of them (respondents were not told that news were false). The list of fake news included misinformation about natural disasters, health, politics and immigration (e.g., “Some vaccines can produce autism in children,” “The President of Venezuela, Nicolás Maduro, called for support of the candidate for the Presidency, Alejandro Guillier”). These so-called news stories had circulated in the authors’ own social media accounts, were fact-checked by El Polígrafo and found to be false. Based on this question we built an exposure variable (range = 0 [not aware of any story] to 14 [aware of all stories]; Cronbach’s α = .70, M = 8.13, SD = 2.76).

Sharing Fake News.

Respondents were also asked if they had shared any of the 14 fake news stories. From their answers we created a variable of sharing fake news (range = 0 [no story was shared] to 14 [shared all stories]; Cronbach’s α = .69, M = .953, SD = 1.50).

Credibility in Fake News.

Subsequently, respondents were asked if they believed in any of the news stories we had shared, regardless of whether they had heard them before or if they had shared them (range = 1 [“not believable at all”] to 5 [“Very credible”]). Based on this question we built a credibility variable of fake news (Cronbach’s α = .85, M = 2.58, SD = 0.67).

2.4 Social Media Use

Frequency of SNS Use.

Participants were asked how much time they spent on four social networks platforms (Facebook, Twitter, Instagram, and WhatsApp) (range = 0 [do not use that social network] to 7 [use more than 6 h per day]; (Cronbach’s α = 0.62, M = 3.45, SD = 1.25).

News Consumption in SNS.

Respondents were asked how many days per week they consumed news on SNS using an 8-point scale, ranged from “I do not see news or do not have news service” to “Every day of the week” (M = .6, SD = 2.57).

2.5 Political Views and Participation

Online Political Participation.

A battery of 8 questions with different examples of political involvement in social networks was used (e.g., “Change profile picture or status in a social network or WhatsApp in support of a political or social cause”). Respondents were asked about the frequency of such activities using a 5-point scale ranging from “Never” to “Always” (Cronbach’s α = .86, M = 1.97, SD = .72).

Strength of Political Views.

We used a 7-point scale for political ideology ranging from “Very left” to “Very right.” To create the variable for strong political identification, the item was folded into a 4-point scale, ranging from weak to strong political views (M = 0.29, SD = 0.45).

2.6 Trust and Physiological Variables

Trust in Traditional Media.

Respondents were asked about 9 statements concerning trust in traditional media (press, radio and television) such as “They are reliable sources of information” or “They present all the sides of a news equally” using a 5-point scale ranging from “Strongly Disagree” to “Strongly Agree” (Cronbach’s α = .86, M = 2.88, SD = .68).

Confidence in Information Shared by Contacts.

Participants were asked about their level of agreement with the statement, “I trust most of the news shared by my social network contacts” using a 5-point scale ranging from “Strongly disagree” to “Strongly agree” (M = 2.31, SD = .83).

Conspiracy Mentality.

Based on Bruder et al. [38], respondents were asked about their level of agreement with 4 conspiracy statements, such as “many very important things happen in the world, which the public is never informed about” and “there are secret organizations that greatly influence political decisions” (Cronbach’s α = .76, M = 3.46, SD = .71).

2.7 Control Variables

Three demographic variables were considered: sex (M = 1.57; SD = .46), age (range = 20 to 71; M = 35.02; SD = 12.32) and educational level (range = 1 [Elementary school incomplete] to 7 [Postgraduate]; M = 5.93, Mdn = 6 [university completed], SD = .72). We also included a variable related to the attention of the respondents to different type of news (Politics, Crime, International news, Economy and business, Food and health, Environment, Technology and Science, Sports and Sports) (range = 1 [no attention to news] to 5 [a lot of attention to news]; Cronbach’s α = .67, M = 3.06, SD = .56). A scale of political ideology ranged from 1, “Very Left”, to 7 “Very Right” was also included (M = 3.8, SD = 1.32). Given that several studies have shown the relevance of online self-efficacy in explaining social media use and behavior [39], we also considered individuals’ perception of their ability to influence their online environment through their skills to understand online information [40]. Thus, participants were asked their level of agreement with 7 statements (e.g., “I can discern between relevant information” or “It is easy to be well informed about important issues”) (range = 1 [Total disagreement] to 5 [Total agreement]) (Cronbach’s α = .74, M = 3.99, SD = .51).

3 Results

3.1 Descriptive Analysis

Before reviewing the proposed hypotheses, we observed the prevalence of exposure, credibility and the propensity to share fake news in the Chilean case using the latest wave of the study, namely 2018. According to table No. 1, we can see the high proportion of familiarity with the fake news revealed in the survey (most respondents had between 30.5% and 84.4% of familiarity). This familiarity is contrasted to fake news about tolerance towards multiculturalism (No. 3, No. 9, and No. 11) and fake news about politics that are not related to the last presidential elections of 2017 (No. 3, No. 10, and No. 2), where participants showed lower levels of familiarity. However, it is important to note that for respondents none of the news stories exceeds 50% of credibility, although 4 of them reach 30% or more. Nevertheless, the highest levels of credibility were found in news about politics and those related to health (No. 6, No. 5 and No. 8). Finally, only a few percentage of respondents said that they had shared fake news. (The range goes from 0.5% to 18.2%.) Thus, except for the news N 4, we see that those fake news stories about politics were less shared than those about health, natural disasters and tolerance to multiculturalism.

3.2 Direct Effects

Returning to our hypothesis, we seek to identify which type of variables are the most relevant in explaining why social media users are exposed, believe, and share fake news through the internet. Figures N1, N2 and N3 presents the SEM models with the direct effects between variables (Fig. 2).

Fig. 2.
figure 2

Direct effects of exposure and credibility of fake news on sharing fake news

Full size image

First, Figure N2 presents the relation between sharing fake news and both variables credibility in fake news and Exposure to fake news. The relation between these variables was tested including the rest of independent and control variables in the model. According with Figure N1 we find a significant and positive relation between sharing fake news and being exposed to fake news (b = .041, p < .01), and sharing fake news stories and believing in them (b = .630, p < .001). This means that the more individuals are exposed to and the more they believe in fake news, the more they will share them in social media (Table 1).

Table 1. Prevalence of misinformation exposure, beliefs, and sharing
Full size table

Regarding the direct effects between the selected variables and credibility of fake news, Fig. 3 shows a negative effect between the frequency of SNS use and credibility in fake news (b = −0.049, p < 0.05). This means that, contrary to what was proposed in hypothesis 1a, the more individuals use the social media, the less they believe in fake news. We also find a positive effect between the confidence in information shared by contacts (b = 0.058, p < 0.5) and holding a conspiracy mentality (b = 0.245, p < 0.01), which allows us to support hypothesis 6a and 7a respectively. We did not find support for hypothesis 2a, 3a, 4a and 5a. This means that in order to explain credibility in fake news, the variables related to trust and physiological factors are more relevant than the political aspects and those related to social media use.

Fig. 3.
figure 3

Direct effects of the hypothesized variables on credibility in fake news and sharing fake news

Full size image

Regarding the role of these variables in exposure to fake news, only online political participation is related significantly to exposure to fake news, which confirms H3b (b = 0.245, p < 0.01). However, none of the others variables have an impact, as Fig. 3 shows.

We also find relevant effects in control variables. Holding strong political beliefs does not seem to explain credibility in fake news, holding more Right-leaning political ideology does have a positive effect in the believing of misinformation in social network (b = 0.071, p < 0.01). This indicates that the orientation of political thought is relevant, with Right-leaning people being more prone to believe in fake news than Left-leaning people. Age also has a significant and negative effects (b = 0.004, p < 0.05), which indicate that older individuals tend to believe less in fake news. Finally, the gender is also relevant with women being more prone to believe in fake news than men (b = 0.089, p < 0.1).

3.3 Mediation Effects

Together with the direct effects described in previous section, we also analyzed indirect effects mediated by the variables “exposure to fake news” and “credibility in fake news.” We found three variables mediated by credibility in fake news: social media use (b = −0.0307, p < 0.5) confidence in information shared by contacts (b = 0.0366, p < 0.5) and conspiracy mentality (b = 0.1542, p < 0.01). These results show that the three variables do have an effect in sharing fake news, but only through the credibility of fake news. So, according to the results, the more the individuals use social platforms, the less they share fake news, mainly because they trust this information less. Similarly, the more individuals hold a conspiracy mentality, the more they share fake news. However, unlike credibility, none of de variables have an indirect effect mediated by exposure.

These results allow us to assume that the variable of credibility in fake news is a relevant variable to explain the indirect relation between several variable and sharing fake news. Also, Fig. 4 shows the importance of trust and psychological variables in explaining the spread of fake news through social media. It again, shows theoretically contradictory evidence about the assumed positive relation between the use of social media and the spread of misinformation through internet (Fig. 5).

Fig. 4.
figure 4

Direct effects of the hypothesized variables on exposure to fake news and sharing fake news

Full size image
Fig. 5.
figure 5

Indirect effects and mediation

Full size image

4 Discussion

Due to its rapid spread in social networks and the harmful effect, the topic of fake news has become a major issue for the public and researchers alike. Justifiable fears exist for the risk fake news poses to democracy, comity, and informed debate. Responsive to this growing concern, we used a panel data to study the topic. To understand the proliferation of fake news in social networks, we compared the effect of different types of factors on three key variables: exposure, credibility and sharing fake news. The findings are discussed below.

First, researchers have consistently argued that the rise of fake news is attributable to the emergence of new digital platforms that facilitate the production and propagation of misinformation. They also suggest that the medium in which public debate takes place makes it difficult for users to discriminate between truthful information and information created for other purposes, including misleading ones. However, our results show that controlling for other variables, the use of social network seems to be negatively associated with believing in fake news. Thus, one possible implication is that more connected users may have developed a sense of awareness about the information quality in social media, and would be less exposed to these types of news.

Second, we outlined that variables related to political identification and participation would have an impact on exposure to fake news, as suggested by previous investigations [11, 16,17,18,19,20,21]. Our results show that individuals who most identify with the political Right are those who are more likely to believe fake news compared to the people identified with the Left. This, however, can be explained by the type of fake news about which the respondents were consulted, which refer to the past government of Michelle Bachelet, president identified with center-left spectrum of politics. In other words, much of the beliefs that researchers themselves hold about who accepts and propagates fake news may be an artifact of the topics chosen as fake news and the specific platforms and aspects they choose to investigate. They would be an ironic and inadvertent demonstration perhaps of Kunda’s earlier findings concerning her cognitive analysis about how people seek to reinforce their prior views and commitments.

Third, as several authors point out, the confidence of users is important in explaining the proliferation of fake news on social networks [17, 21, 24]. Coherently, we find that both confidence in information shared by contacts and conspiracy mentality are important variables to understand why people believe in fake news. However, the relevance of these variables was not significant in predicting exposure to fake news. A plausible explanation could be that trust in contacts and holding a conspiracy mentality make people more susceptible to believe in fake news, but that does not translate into actively seeking out misinformation or engaging with sources that promote them.

And this finding is relevant for two reasons. On the one hand, the fact that users trust the information shared by their contacts may enhance filter bubbles or echo chambers within platforms such as Twitter or Facebook, either by the algorithms of these networks or by a natural tendency of groupings based on similar interests. This aspect is central, because it may augment the disinformation effects of the fake news, since it prevents users engaging with others who will present them with counter-acting dissonant information.

On the other hand, we also find support for the idea that there are certain types of individuals who, due to personality traits (e.g., lack of confidence, paranoia, low self-esteem), constitute ideal victims of fake news. This was shown in the relationship between conspiracy mentality and believing in fake news.

Finally, we found three relevant indirect effects mediated by credibility in fake news, which are social media use, conspiracy mentality and confidence in formation shared by contacts. The presence of this indirect effect underscores the need to analyze with more detail and complexity the relation between different factors that influence misinformation in internet and social media. The indirect effects found in this research allow us to asseverate that using social media, holding a conspiracy mentality or trust in information shared by contacts are not enough on their own to explain why people spread misinformation through internet. Even in the presence of these variable, if individuals do not believe in the information to which they are exposed, based on our data, we cannot expect that they will share these pieces of information.

In short, the literature on fake news has provocatively addressed several theories. In this study, we have tried to take the most important variables offered by the literature and observe how they operate when they are included within a single integrated and longitudinal model. Our intention was to advance our understanding of a long-standing phenomenon, such as fake news, which today has commanded still greater attention due to social media. We know much about how each dimension works separately, but in literature there is a lack of analyses that integrate different dimensions. Our effort here is to respond to this vacuum by conducting (what we believe to be) the first longitudinal study in the area.

However, much remains to be done. In this article we analyze the case of a Latin American country, Chile. Even though many of our conclusions may be useful for current debates among scholars, there may be no reason to expect that other region around the globe behave in the same way regardless misinformation passing through the internet. However, it is relevant to note that one of the limitations of this study is the fact that more than half of the users surveyed in the first wave were not included in the second wave. One reason that may explain this lower participation is the time elapsed between the first and second waves (15 months). Another limitation is related to the fact that we analyzed several fake news stories without considering if they refer to political, health, migration or tolerance themes. Further investigation can separate fake news stories based on topics and may find different patterns between variables. In spite of the limitations, we hope that this article could help as an initial step for more comprehensive and multidimensional studies about fake news and misinformation in the internet.