As argued above, social media has displaced traditional media channels as a primary gateway for the public to access scientific information. On the one hand, a majority of Americans report incidental exposure and attention to science news (Funk et al., 2017; Hitlin & Olmstead, 2018), as well as sharing such information on social media channels such as Twitter, Facebook, and YouTube (Hargittai et al., 2018). Social media use is associated with public trust in science (Huber et al., 2019) and knowledge gain (Huber et al., 2019; e.g., Mueller-Herbst et al., 2020). On the other hand, scientists and scientific institutions can directly communicate to the public through social media with no intermediate gateways, resulting in “the most genuine type of communication and thereby allowing scientists to fulfill their accounting duty effectively by engaging eye to eye with the general public” (Weingart & Guenther, 2016, p. 5). Indeed, experimental studies show that the public doesn’t discount the credibility of scientists—often cited as a reason for scientists’ lack of engagement—even when they advocate on contested issues (Kotcher et al., 2017). Experts can also correct misinformation online (Bode & Vraga, 2015; Vraga & Bode, 2017).

The primary pathway of knowledge gain from social media is through incidental and habitual exposure to issues (Baum, 2002; Bode, 2016; Cacciatore et al., 2018). Research on the potential of incidental learning through passive consumption of mass media is well documented in political communication research (Krugman & Hartley, 1970) and has been shown to be associated with public perceptions of science and technology issues (Nisbet et al., 2016; Nisbet & Markowitz, 2016). Termed as a “gateway effect,” this process explains that as individuals encounter “hard news” alongside more entertaining content, they are more likely to learn and become engaged with various issues. Importantly, as such exposure occurs without activating sensors for motivated or purposive information gain, it may result in knowledge gain particularly among less attentive individuals (Krugman & Hartley, 1970; Tewksbury et al., 2001). Both forms of incidental exposure and active news seeking are related to engagement with news content (Oeldorf-Hirsch, 2018).

In the few studies on social media use and scientific knowledge gain, mixed results are common. Based on a cross-country sample, Huber et al., (2019) found that social media news use media was positively associated with scientific knowledge, and had a stronger association than traditional news use. Based on a national sample, Su et al., (2015) found that people who prefer to primarily rely on online-only sources for science news have a significantly higher level of knowledge of science issues than those who primarily rely on traditional formats for science news.

However, other studies have found that social media use was associated with less awareness and knowledge. Mueller-Herbst (2020) found that Facebook use was negatively associated with gene editing awareness. Instead, they found time spent on Facebook was positively associated with awareness, arguing that more involvement results in higher chance of exposure to scientific news topics and therefore knowledge gain. In a recent study, Stecula et al., (2020) found that compared to traditional media uses, social media users were more likely to be misinformed about vaccines. Bridgman et al., (2020) found that exposure to social media was associated with misconceptions about the COVID-10, which in turn was associated with lower compliance with social distancing measures in Canada.

Social media contains a fair amount of misinformation (e.g., Bridgman et al., 2020, Scheufele & Krause, 2019; Wang et al., 2019) and exposure to inaccurate or fake information can diminish public trust in science and public understanding of scientific issues. For example, several studies in climate change communication show that exposure to inaccurate coverage likely results in beliefs inconsistent with scientific consensus about climate change (e.g., Feldman et al., 2012; Zhao et al., 2011). In other words, use of social media can potentially lead to either increase or decrease in scientifically consistent beliefs and knowledge based on the exposure, attention, and elaboration of information one sees on social media.

This mixed finding between social media use and knowledge gain about scientific issues deserves more attention. We extend the above studies—primarily based on exposure or use of social media—by focusing on the role of trust as a more proximate source of influence on knowledge gain. A number of studies on risk perceptions and science communication highlight that trust serves as a heuristic, particularly on complex scientific and health topics.

While trust has been a central component is several science communication studies, there is a “conceptual confusion” (Lewis & Weigert, 1985, p. 975) about the definition of trust (see Schäfer, 2016; Hendriks et al., 2015). Here, we adopt the definition of trust as commonly accepted (Earle, 2010; Poortinga & Pidgeon, 2003), as “a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behavior of another” (Rousseau et al., 1998, p. 395). In line with recent conceptualization and measurement of trust (Hendriks et al., 2015), it is inferred as comprising expertise, integrity, and benevolence. Particularly, media as trust intermediary between scientists and public deserves our attention as public primarily comes to know about scientific breakthroughs through media. As Schäfer (2016) argued, trust in science and trust in media as a source of information has to be distinguished to “enable researchers to assess the relative importance of both factors in the production of mediated trust in science” (p. 3).

A large and substantial body of research elucidates that for individuals, information from a trusted source helps reduce cognitive effort and provides a short-cut to make judgements about an issue (e.g., Hmielowski et al., 2013; Malka et al., 2009; Poortinga & Pidgeon, 2003; Siegrist et al., 2005; Slovic, 1993). Individuals are cognitive misers (Fiske & Taylor, 1991) and they may not either have the required time or cognitive ability to understand any one particular issue comprehensively, especially for a complex and emerging pandemic such as the coronavirus that continues to challenge scientists.

A number of studies show that trust in scientists and scientific institutions is one of the strongest predictors of knowledge and general attitudes towards science (e.g., Nisbet & Markowitz, 2016; Stecula et al., 2020). Recently, based on a nationally representative sample of Americans, Stecula et al., (2020) found that distrust of medical authorities was one of the strongest predictor of vaccine misinformation, and this was true across different demographic groups and political beliefs. These associations were resilient over time even with increasing media coverage of CDC and NIH following measles outbreak in the US in 2019. At the same time, scientific advances—such as discovery of vaccine for Zika virus—help increase public trust in science (Hilgard & Jamieson, 2017).

Trust in social media channels may function in a similar way as trust in traditional media and scientific organizations, in that, people who trust these channels are more likely to frequently use them, resulting in repeated exposure that may result in knowledge gain (Bode, 2016; Cacciatore et al., 2018). Moreover, due to the affordances of social media such as liking, sharing, commenting, social media may result in stronger attitudes and beliefs about scientific issues, thereby increasing their attitude certainty and knowledge (Media Insight Project, 2017). Trust in social media can result in either heuristic judgments—incidental exposure—or result in elaboration and discussion through commenting and sharing information with one’s social networks (Ho et al., 2017). To our knowledge, no previous study has evaluated trust in social media sources on shaping public knowledge about scientific issues. We test the relationship by controlling a series of factors that have shown to result in knowledge gain, such as attention to news (e.g., Oschatz et al., 2019), trust in traditional media such as newspapers (e.g., Huber et al., 2019; Stecula et al., 2020), and trust in scientific organizations such as CDC and NIH (e.g., Drummond & Fischhoff, 2017). In addition, a risk perception variable—perceived timing of harm—was also used as a control variable.

A number of studies report a consistent and emerging polarization along political lines on scientific issues in the US (Dunlap et al., 2016; Feldman et al., 2012; Kahan et al., 2012). Such polarization, worryingly, increases with increase in education and scientific knowledge levels, on a variety of scientific topics such as stem cell research, genetically modified crops, climate change, among others (Bail et al., 2018; Drummond & Fischhoff, 2017; Kahan et al., 2012). Opinion polarization along political lines is a result of partisan media exposure and motivated reasoning. Diehl et al., (2019), for example, found that the association between social news media use and climate change belief was moderated by conservative political ideology. Partisan polarization can increase in relatively short time; Gustafson et al., (2019) found that in just four months, the divide between Democrats and Republicans in support of Green New Deal doubled from 33 percentage points difference in December 2018 to 64 percentage points in April 2019. Myers et al., (2017) found that political ideology has a strong influence on public trust in federal scientific research. While previously untested, it is important to understand if the potential knowledge gain using social media is also aligned along partisan lines. As there are no studies guiding development of hypothesis specific to trust in social media—and the mixed findings of use and social media news use—we seek to answer the following research questions:

RQ1: How is trust in social media as source of accurate information about the coronavirus associated with knowledge about the coronavirus?

RQ2: Are there any partisan differences in the relationship between trust in social media and knowledge about the coronavirus?

The dependent variable of scientifically accurate knowledge was measured by asking a series of fifteen questions in the following way: “To the best of your knowledge, are each of the following statements true or false.” The response options were, Yes (1) and No (0); The response categories of “Don’t know” and those who refused to answer were coded as missing data. Eight statements were scientifically accurate relating to symptoms (dry cough (1, Yes, 82%), fever (90%)), spread (“the coronavirus can be spread by people who do not show symptoms,” 91%), “the coronavirus can live on some surfaces for up to three days,” 78%), protection (frequent hand washing (92%), avoiding large gatherings (92%), maintaining 6-feet distance physical distance with others (74%)), and cure (“there is currently no cure for the coronavirus (81%). Five statements were scientifically inaccurate and refereed to a variety of misconceptions including, “Drinking water can flush the coronavirus into your stomach where the acid will kill it (18%),” “antibiotics can prevent or kill the coronavirus (12%),” “there is currently a vaccine available to the public that is proven to prevent coronavirus infection (10%),” “sneezing is a symptom of coronavirus infection” (39%)”, “hand sanitizers are better than soap and water for preventing infections (24%).” Two questions about non-contagion (28%) and immunity (25%) after recovery were not used in the analysis as currently there is no evidence to state either according to WHO, even as these were covered broadly in the media. The dependent variable was computed by adding all the scientifically accurate statements and then subtracting them from the responses to statements inconsistent with WHO information about the COVID-19 (M=5.86, SD=1.98) KR-20 (Kuder-Richardson Formula 20) = .46).

Trust in information from social media channels was gauged by asking respondents, “How much they trust or distrust the following as a source of accurate information about the coronavirus?” The responses ranged from “Strongly distrust” to “Strongly trust.” The “Don’t Know” responses were coded as the middle category, while if the respondent choose not to answer (only three respondents) were coded as missing data. Similarly trust in information from national newspapers (“such as New York Times, Washington Post, USA Today, etc”) was coded on a five-point scale (M=3.53, SD=1.30).

Similarly, trust in scientific institutions was measured on a four-point scale by asking respondents about how much they trust or distrust the following as a source of accurate information: CDC (M=3.35, SD=.76) and NIH (M=3.25, SD=.75).

Attention to the news (M=3.56, SD=.63) was measured by asking respondents, “How closely have you been following news about the coronavirus?” The responses were measured on a four-point scale from “Not at all closely” (1) to “Very closely” (4); Only three respondents who refused to answer the question and were treated as missing data.

Perceived harm was measured by asking respondents, “When do you think the coronavirus will start to harm people in your community?” (M=5.03, SD=1.45). The response options were: “Never” (7%), “More than 3 months from now” (2.4%), “In 2 to 3 months” (3.2%), “In about a month” (9.7%), “In the next 2 to 3 weeks” (23.1%), and “They are being harmed right now” (54.5%). Only one respondent did not answer the question.

Demographic variables, listed in Table 1. Income was measured by asking respondents to estimate their total household income in the past 12 months before taxes: Less than $25,000 (19.8%), $25,000 to $49, 999 (20.9%), $50,000 to $74,999 (18.6%), $75,000 to $99,999 (11%), $100,000 to $149,999 (20.3%), and $150,000 or more (9.4%). Ethnicity was dummy coded to compare those who self-identify as White (62%) compared to others. Region was dummy coded comparing individuals’ living in South to living in Northeast (17%), Midwest (21%), and West (24%). Political party affiliation (M=3.93, SD=1.80) were measured using 7 categories (very liberal Democrat, somewhat liberal Democrat, moderate or conservative Democrat, Independent or non-leaning, liberal or moderate Republican, somewhat conservative Republican, very conservative Republican). Those who did not respond or mentioned no party or not interested in politics were recoded as Independents/non-leaning.

Analysis proceeded in a step-wise fashion by examining bivariate relationships and testing a series of regression models. Moreover, political party identification were coded in multiple ways to test the robustness of the interaction results, including interactions with dummy coded variables. Hayes (2013) SPSS PROCESS macro was used to test interactions between political identity affiliation and trust in social media sources. The bootstrap analysis was conducted with 10,000 iterations and bias-corrected estimates.

A primary limitation is that it is a cross-sectional study and cannot determine if trust in social media resulted in lack of knowledge or people who already have strong misconceptions about the coronavirus—particularly those who mistrust traditional media—tend to trust social media. Our analysis relies on a series of Yes or No questions, similar to True/False questions often used to assess knowledge (Su et al., 2017). A more comprehensive assessment of public information levels are necessary. While the measures of trust were specifically related to accuracy of information about the coronavirus on social media, more refined measures of specific kinds of trust in information on social media will give a comprehensive picture. For example, trust can be measured based on the dimensions of expertise, integrity, and benevolence (Hendriks et al., 2015), even as several studies use just one overall measure of trust. It is unclear why trust in national newspapers was not associated with knowledge. It could be because of a general decline of trust in newspapers or increasing skepticism particularly related to coverage of science news coverage. The intense opinion polarization along political lines is particularly visible in the US and may not be applicable in to other countries. Nevertheless, cultural factors such as individualism and egalitarianism—not considered in this study—have been found to be associated with political partisanship and shown to influence risk perceptions worldwide. Furthermore, the study did not examine any of the content of the science information encountered online. In fact, it may be impossible to curate each individuals’ social media feed to make any conclusive arguments about exposure to kinds of information people encounter on social media. While we were able to detect partisan effects for Republicans who trust Twitter, it is unclear why such an affect was not observed either for Facebook or YouTube, or for Democrats, even as other research indicates that social media driven polarization is particularly high for Republicans (Bail et al., 2018). Future research should further test the impact of network heterogeneity in impacting knowledge gain through trust in social media channels.

Social media is now the primary way through which public access news about science and health issues, even as traditional science journalism is in decline. As a result, trust in social media as an accurate source of information is important for public knowledge and engagement with complex and emerging scientific issues. The findings of this paper are troublesome—trust in social media channels is negatively associated with basic knowledge about the coronavirus even as social media organizations have taken a war effort to control the spread of fake news and misinformation. Our ability to contain the “infodemic” will be also be our key to saving lives due to the coronavirus pandemic.