Avoiding Covid-19 risk information in the United States: The role of attitudes, norms, affect, social dominance orientations, and perceived trustworthiness of scientists.

This study seeks guidance from the planned risk information avoidance model to explore drivers of risk information avoidance in the context of COVID-19. Data were collected early during the pandemic. Among our most notable results is that participants who are more oriented toward social dominance and are more skeptical of scientists' credibility have (1) more supportive attitudes toward risk information avoidance and (2) feel social pressure to avoid risk information. The findings of this study highlight how the role of skepticism in science and intergroup ideologies, such as social dominance, can have important implications for how people learn about health-related information, even in times of heightened crisis.

INTRODUCTION

Almost immediately following its discovery, news coverage of the virus that causes COVID‐19 (sars‐cov‐2) became ubiquitous across all news channels in the United States. All aspects of American life were discussed within the context of the virus, including healthcare, politics, manufacturing, energy, agriculture, finance, and education (Basch et al., 2020). The continuous news cycle, coupled with peoples’ lived experiences during COVID‐19, contributed to an ongoing state of information overload and prolonged emotional stress for many individuals (Fiorillo & Gorwood, 2020; Garfin et al., 2020; Kim & Kreps, 2020; Valika et al., 2020).

Not all of the available information has been helpful, as it has included contradictory and even inaccurate information (Parmet & Paul, 2020). Indeed, the misinformation littering the COVID‐19 information landscape has worked against public health efforts to slow the spread of the virus (Kim et al., 2020; Tasnim et al., 2020) and has been associated with negative individual health outcomes (De Coninck et al., 2021). To make matters worse, as the public sought information about how to avoid contact with the virus early on, even information from the U.S. government and the World Health Organization (WHO) contained contradictions about the efficacy of simple precautionary behaviors such as mask wearing (Parmet & Paul, 2020; Viera et al., 2020).

The information overload and regular stream of misinformation led some individuals to avoid information about the risks associated with COVID‐19, as a way to cope with the stress (Kim et al., 2020; Soroya et al., 2021). However, while avoiding information is a viable coping mechanism for reducing stress (Cheng et al., 2021) and has been explored as a productive situational coping strategy (Ytre‐Arne & Moe, 2021), it may have implications for controlling the community spread of the virus. That is, while risk information avoidance (RIA) may serve as a coping mechanism at the individual level (Link, 2021), at the societal level it may thwart efforts to disseminate public health information that can mitigate community spread and save lives (Howell & Shepperd, 2013; Siebenhaar et al., 2020). By late spring 2020, a Pew study on Americans’ information behaviors found that about seven in 10 adults reported they needed to take breaks from COVID‐19 news (Mitchell et al., 2020). As a result, more research is needed to create effective information campaigns that can respond to individuals’ information management needs, while also facilitating their engagement with vital public health messaging (Cheng et al., 2021; De Coninck et al., 2020).

To further explore the public's risk information avoidance behaviors during the pandemic, we sought guidance from the planned risk information avoidance framework (PRIA; Deline & Kahlor, 2019). According to the PRIA model, which is explained in more detail below, COVID‐19 risk information avoidance is likely driven by risk‐related cognitive, affective, and socio‐cultural factors. For example, Deline and Kahlor (2019) explain that “an injunctive avoidance norm [one of the socio‐cultural factors in PRIA] could entail an implicit group rule that if members avoid information about the potential [risk] they will be positively judged” (p. 367). However, given the pronounced role of misinformation in the pandemic information landscape, we further propose that two other constructs are likely to play a role in COVID‐19 risk information avoidance in the United States: perceived trustworthiness of scientists and social dominance orientation (SDO). These additional constructs address phenomena that have seemingly driven health prevention behaviors in the United States: skepticism in science (Plohl & Musil, 2021) and political tribalism (de Bruin et al., 2020).

When faced with new scientific information, such as a novel coronavirus, many people lack a readily accessible cognitive framework with which to interpret the information. As a result, they may instead process the information using heuristics such as rapid judgements about the credibility of information sources or interpretive cues drawn from their own attitudes or ideologies. Early on during the pandemic, scientists and public health experts were an essential information source for pandemic‐related risk communication, and they were also the most highly trusted (McFadden et al., 2020). However, their perceived trustworthiness was undermined over time by politicians and conservative media commentators who cast doubt upon scientists and questioned whether they should be trusted as a source of information about the pandemic (Jamieson & Albarracin, 2020; Plohl & Musil, 2021; Romer & Jamieson, 2020).

Moreover, early on, as the pandemic evolved, a sort of political tribalism emerged in the United States related to the public's views about the virus and how to best handle it as a nation (de Bruin et al., 2020; Druckman et al., 2020). For example, in one study, media use contributed to “the partisan divide in preferences for the timing of lifting economic restrictions and reported protective behaviors” (de Bruin et al., 2020, p. 177). This observed polarization is an intergroup phenomenon that impacts both attitudes and behaviors (beyond political party identification), which is why we selected social dominance orientation as a framework for exploring the division that surfaced during the pandemic. Social dominance orientation may have played a role in driving risk information avoidance related to COVID‐19 (Ho et al., 2012).

LITERATURE REVIEW

Risk information avoidance

Sweeny et al. (2010) explicate information avoidance as a coping behavior that is intended to “prevent or delay the acquisition of available but potentially unwanted information” (p. 341). They assert that individuals are more likely to engage in avoidance behaviors when they are presented with information that may lead to an unwanted change in their beliefs or behaviors, or cause an emotional burden. The ability to cope with or handle potential changes in beliefs, actions, or emotions is situational, and extraneous stressors can have an impact on coping strategies. In the context of COVID‐19, individuals were inundated with information about the virus and how the pandemic was impacting all aspects of American life: an economic slowdown, the shutting down of the nation's schools and childcare facilities, mounting death rates, the fracturing of public opinion about how best to deal with the virus, and more. For some individuals, the overall uncertainty of the situation combined with the sheer number of areas in which uncertainty prevailed could have easily motivated risk information avoidance (Sorrentino & Hewitt, 1984).

Avoiding information can have both positive and negative consequences. On the one hand, risk information avoidance can mitigate the overwhelming stress that comes from information overload; too much information can lead to depression and anxiety (Siebenhaar et al., 2020). On the other hand, from a disease control standpoint, avoiding risk information can have negative consequences, which in this case includes the unmitigated spread of disease within some communities with a higher incidence of information avoidance (Howell et al., 2014). For example, in communities with lower risk mitigation resources and a lack of community support, risk information avoidance can proliferate—there is no point in exposing oneself to disease prevention information when the prevention practices are not possible within one's community (Howell et al., 2014). There is evidence to suggest that testing for disease (e.g., sexually transmitted diseases or COVID) is sometimes avoided due to the fear of receiving a positive diagnosis and not knowing how to cope with that outcome or deal with negative social sanctions due to social stigma (Earnshaw et al., 2020), however, avoiding a COVID‐19 test due to the fear of a positive diagnosis poses a threat to other individuals, because the disease carrier may fail to prevent the potential spreading of the disease if they are not aware of their diagnosis (Caplin & Eliaz, 2003).

Looking at information exposure specifically, research suggests that individuals who were overwhelmed and confused by COVID‐19 information were more likely to skim the available information rather than process it thoroughly, thereby decreasing their exposure to information about disease mitigation (Hong & Kim, 2020). Furthermore, Siebenhaar and colleagues (2020) found that distress was related to (risk) information avoidance, which in turn reduced compliance with preventive measures such as social distancing and staying home. These studies suggest that a portion of the population is not being fully exposed to key COVID‐19 information, including local infection rates, opportunities for getting tested, and evolving vaccination opportunities, all of which are key for curbing the pandemic (Schmid et al., 2017). This research suggests that a more nuanced understanding of the cognitive, affective, and sociocultural factors that contribute to COVID risk information avoidance may lead to more responsive and effective messaging strategies.

In the burgeoning research on risk information avoidance, such avoidance has been characterized in sometimes conflicting ways—as both an active and passive behavior, as both self‐ and other‐oriented, and as a reaction to both negative and positive risk information, as well as known and unknown risk information (Deline & Kahlor, 2019; Kahlor et al., 2020; Narayan et al., 2011; Sweeny et al., 2010; Sweeny & Miller, 2012; Yang & Kahlor, 2013). In this study, we conceptualize risk information avoidance as the active avoidance of information about a given risk, and a common behavior in which people actively avoid hearing or seeing risk‐related information. Such behavior can include shutting off a radio or television, leaving a Web page, or changing the topic of a conversation (Barbour et al., 2012; Narayan et al., 2011). We further conceptualize risk information avoidance as an information behavior that is distinct from information selection and seeking (Barbour et al., 2012; Narayan et al., 2011), and inertia, which refers to a lack of action to support the status quo (Polites & Karahanna, 2012).

In an effort to build on the work of Sweeny and others, and integrate the myriad concepts that have surfaced as relevant for risk information avoidance behaviors, Deline and Kahlor (2019) proposed the PRIA framework. The framework emphasizes the deliberate and active nature of risk information avoidance and accounts for the social, psychological, and affective constructs that undergird the behavior. To our knowledge, PRIA is the most comprehensive (and broadly applicable) theoretical model for exploring risk information avoidance. However, it is important to note that PRIA is a “‘pre‐theoretical conceptual representation’ of a phenomenon…that can be used to elaborate theory, integrate research fields and test variable relationships” (Deline & Kahlor, 2019, p. 376). The intention of the researchers in developing PRIA was to remind researchers of factors that should be the focus of future investigations (Deline & Kahlor, 2019; McPhee & Poole, 2016).

With that in mind, we sought guidance from the PRIA framework when selecting which concepts to include in our study of risk information avoidance in the context of the COVID‐19 pandemic. Specifically, the PRIA model recommends the inclusion of cognitive characteristics—we chose perceptions of the risk and attitudes toward avoiding COVID‐19 risk information, affective factors—we chose affective response to the risk, and sociocultural factors—we chose subjective norms related to avoiding COVID‐19 risk information. These concepts have been robust contributors to risk information behaviors in other COVID‐19 studies (Ahn et al., 2021; Kim et al., 2020; Moon et al., 2021).

However, building further on the theoretical underpinnings of PRIA, and recognizing that misinformation and politicization have proliferated during the pandemic, we also chose to include two additional factors that are associated with the desire to maintain a consistent self‐concept (Narayan et al., 2011) and manage uncertainty (Barbour et al., 2012). These factors are social dominance orientation, which is an underpinning of political ideology that captures the tendency to favor social hierarchies and inequalities, and perceived trustworthiness of scientists, both of which are reflections of what Sweeny and colleagues (2010) call certainty orientations (see Figure 1).

FIGURE 1

The planned risk information avoidance model Note: Bolded concepts are variables of interest in this study; Grey concepts are variables proposed in the PRIA but not included in this study as PRIA was developed by Deline and Kahlor (2019) to scaffold further theorizing, as we do here; italicized concepts are additions to the original theorization of the PRIA model.

For example, recent research suggests that a social dominance orientation is correlated with the sharing of misinformation (Lobato et al., 2020). Earlier research also suggests that SDO, which emphasizes support for hierarchical structures and leadership, could motivate individuals to avoid information that challenges or questions the leaders they support (Sweeny & Gruber, 1984). In the case of science mistrust, SDO has been shown to be a negative predictor of perceptions of scientists’ credibility and is associated with the rejection of mainstream science and scientists in the context of climate change, genetically modified foods, vaccines, and fluoridation (Kerr & Wilson, 2021). High SDO people are likely to be cynical of science and scientists because most of it is seen as taking place in universities, which are seen by some as hierarchy‐attenuating institutions that serve to decrease inequalities (Kerr & Wilson, 2021). Looking at trustworthiness of scientists specifically, Van Mulukom (2020) found it related to adherence to “scientifically supported, government‐recommended protective guidelines” (p. 2). Van Mulukom further describes pandemic‐related behaviors as the “result of complex clusters of variables involving sources [authorities and their media], information, and concern and perceived risk” (2020, p. 2), which further recommends the inclusion of social dominance orientation and perceived trustworthiness of scientists with PRIA as we explore COVID‐19 risk information avoidance.

Attitude toward avoidance

The inclusion of attitudes is not new when theorizing what risk information avoidance is related to. Attitudes toward behaviors, including risk information avoidance, are theoretically rooted in the Theory of Planned Behavior (Ajzen, 1991) and are considered positive cognitive predictors of avoidance that capture both individuals’ instrumental beliefs (akin to a utility evaluation, as it is useful to avoid information) and experiential beliefs (akin to an emotional evaluation, as it feels appropriate to avoid information) about behavior. Looking at risk information behaviors broadly, if individuals hold a positive outcome expectation for seeking or avoiding risk information, they are more willing to engage in the respective behavior of seeking or avoiding (Ho et al., 2014; Hovick et al., 2014; Kahlor, 2010; Kahlor et al., 2020). For example, Kahlor and colleagues (2020) found that a favorable attitude toward carbon capture and storage risk and benefit information avoidance positively predicted risk information avoidance behavioral intentions. Therefore, we propose:

An individual's attitudes toward avoiding information will be positively related to their COVID‐19 risk information avoidance.

Perhaps even more interesting than the relationship between information‐related attitudes and information behaviors are the other cognitive factors (concurrent perceptions, attitudes, and beliefs) that surface when one considers whether to engage in risk information behaviors. Those other factors include risk perception, affective response to risk and trustworthiness of scientists.

Risk perception

Another cognitive component cited in PRIA is risk perception, which represents how individuals perceive and evaluate risk. Although it is a multidimensional concept, literature often focuses on risk perception as the perceived likelihood and severity of risk (Deline & Kahlor, 2019; Kasperson et al., 1988). PRIA adopts a similar conceptualization to reflect individuals’ cognitive evaluations of risk, and Deline and Kahlor (2019) argue for a quadratic relationship between risk perception and risk information avoidance, suggesting that the effects vary by level. Given the nature of risk information avoidance as a stress coping strategy, a high risk‐perception level is likely to activate avoidance. At the same time, low risk estimation can raise interest in the topic and motivate information monitoring behaviors. However, similar to the risk information management models upon which PRIA is built (RISP, Griffin et al., 1999; PRISM, Kahlor, 2010), we propose that risk perception impacts risk information avoidance through affective response to the risk.

Affective risk response

When discussing risk perception, it is hard to discuss it alone without affect or emotion, as they are often strongly correlated within empirical research (Griffin et al., 2004). Previous risk information avoidance research also reported this relation and further showed that risk perception's impact on risk information avoidance is an indirect one, mediated by affective risk response (Kahlor et al., 2020; Yang & Kahlor, 2013). According to Slovic (1993), when most people evaluate a risk, they may consider available heuristic cues and think more analytically about the risk. Affective cues, especially negative affective cues, such as dread and worry, are among the heuristic cues that contribute to people's evaluation of risk (Slovic, 1997). That is, does the risk make people feel a sense of dread or worry? Oftentimes, these negative affective responses to a risk are tied to such characteristics as fatalities, catastrophes, and uncontrollability.

Given the context of COVID‐19, including the mortality rate and overall uncertainty of the pandemic, we similarly focus on the relationship between risk perceptions and affective risk response (Griffin et al., 2013; Kahlor, 2010). Research suggests that negative emotions can have a divergent impact on risk information avoidance, depending on the intensity of the emotion felt. For example, people may avoid risk information when they perceive a low level of behavioral control (see Witte, 1992). In other cases, negative emotions can drive an individual's attention to additional information and decrease risk information avoidance (see Kahlor et al., 2020; Yang & Kahlor, 2013). However, in the case of COVID‐19 specifically, Kahlor et al. (2020) suggest that fear, anger, and hope are positively related to risk information avoidance. As a result of the extant research on risk perceptions and risk‐related affect, our study proposes the following hypotheses:

An individual's risk perceptions related to COVID‐19 will be positively related to affective risk response.

An individual's affective risk response will be positively related to risk information avoidance in the context of COVID‐19.

Perceived trustworthiness of scientists

In addition to the more common aspects of risk perception detailed above, such as likelihood and susceptibility, risk perceptions are generally conceptualized as much more multifaceted and often include other dimensions including trust in the institutions that are perceived as playing a role in the risk situation (Griffin et al., 1999). These institutions may be the creators of the risk, they may simply offer expertise about the risk, or they may be potential mitigators of the risk. Such institutions may include scientists, public health officials, private industries, etc. As COVID‐19 emerged in the United States, scientific experts from the U.S. National Institute of Allergy and Infectious Diseases and the Centers for Disease Control emerged as public information sources for explaining the science of the pandemic and how to protect communities and individuals from the virus. As a result, individuals formed judgments about these experts. According to Wynne (2006), included in these judgments is the extent to which individuals feel they can trust the experts to protect them from harm.

Given the novel nature of the pandemic and the virus at its root, we were most interested in perceptions of scientists as sources for risk information, specifically, we sought to explore individuals’ perceived trustworthiness of scientists. We anticipate that trustworthiness of scientists will be highly correlated with risk perceptions given that, at the early stages of the pandemic, national scientific experts had a (mostly) unified voice suggesting that the virus was extremely contagious and that Americans should practice social distancing and self‐isolation to avoid contact with it. Given research suggesting a positive relationship between trustworthiness of scientists and risk perceptions (Plohl & Musil, 2021), we expect:

Perceived trustworthiness of scientists will be positively related to COVID‐19 risk perceptions.

Our conceptualization of the perceived trustworthiness of scientists refers to the individual's evaluation of scientists’ expertise, integrity, and benevolence (Hendriks et al., 2015). Such perceived trustworthiness can impact risk information avoidance in two main ways: first it can impact an individual's evaluation of the risk itself (as we suggest above) and second it can impact the individual's attitudes toward avoiding the available information. According to Slovic (1993), the perceived trustworthiness of risk management agents plays a vital role in an individual's risk evaluation, because trust in the management agents impacts one's estimation of the predictability of the risk (which affects risk perception) and its controllability (which affects affective risk responses). The more one mistrusts the agent, the more one mistrusts that agent's information related to the risk. Scientists, primarily epidemiologists and medical scientists, have been essential agents of pandemic risk management. Throughout the pandemic, risk management has relied heavily on these experts to supply knowledge, provide instructions, and seek solutions.

Meanwhile, partisan rhetoric and disdain for scientists has emerged to cast doubt on the scientists and the science (Jamieson & Albarracin, 2020; Plohl & Musil, 2021; Romer & Jamieson, 2020). The disdain and doubt made information seeking more challenging and unpleasant, because scientists’ credibility no longer served as an effective heuristic cue when judging information quality (Hendriks et al., 2015). One potential impact of this disdain and doubt was that individuals faced the information hodgepodge with added uncertainty about whether it was necessary to find an alternate source of information aside from scientists. This additional uncertainty added to an individual's cognitive load when faced with COVID information. In such circumstances, individuals may cope by considering risk information avoidance as more desirable than seeking (or engaging with) risk information (Sweeny et al., 2010). To this end, we anticipate that the perceived trustworthiness of scientists will be related to attitudes toward risk information avoidance such that:

The less likely an individual is to perceive scientists as trustworthy, the more positive their attitudes toward risk information avoidance related to COVID‐19.

When faced with uncertainty and information overload, another coping mechanism is to rely on social norms as a processing heuristic; such a heuristic may cue individuals to seek or avoid information depending on their social frame of reference (Smith, 2006). Social norms are perceived patterns in social behavior (more on norms below) that provide contextual cues to guide behavior (Bicchieri, 2005; Colombo, 2014). In the pandemic context, when it becomes difficult to distinguish authentic information from false narratives, people may seek guidance from their social groups and thus become more sensitive to social norms as heuristic information processing cues. Thus, we propose the following:

As the perceived trustworthiness of scientists decreases, the more an individual will adhere to subjective information avoidance norms.

Social dominance orientation

Social dominance orientation is another social‐cultural factor we are interested in exploring in this work. SDO describes an individual's preference for group‐based inequality and dominance versus equality and inclusion and is “the extent to which one desires that one's in‐group dominate and be superior to outgroups” (Pratto et al., 1994, p.742). Thus, individuals who are high on SDO see themselves as members of the dominant group, preferring ideologies and policies that reinforce hierarchical relationships and maintain their position of privilege within it. Given the partisan information landscape for COVID‐19, we sought potential explanations for intergroup ideologies, social political attitudes, and prejudice (Jamieson & Albarracin, 2020; Plohl & Musil, 2021; Romer & Jamieson, 2020; Pratto & Stewart, 2012). The literature on SDO suggests that the orientation toward social dominance drives a compulsion to engage in intergroup ideologies that increase intergroup inequality (i.e., hierarchy‐enhancing ideologies such as racism) for those with high SDO, or decrease intergroup inequality (i.e., hierarchy‐attenuating ideologies such as feminism) for those with low SDO (Pratto & Stewart, 2012).

Although SDO shares some conceptual overlap with cultural cognition theory, it is distinct. In the context of science communication, cultural cognition offers an explanation for why individuals disagree about social risks like climate change and focuses on information seeking and avoiding as a way for individuals to reinforce in‐group social ties (Kahan, 2010). SDO, on the other hand, goes beyond risk perceptions to explain broader patterns of (in)equality and why individuals act to preserve and promote their own group dominance. While cultural cognition and SDO both offer explanations for how groups promote internal cohesion, SDO also is outward facing, explaining why individuals are prejudiced against others or support out‐group inequality (Santos & Feygina, 2017).

The strong compulsion toward an intergroup ideology also suggests that SDO is strongly correlated with social norms. Because high SDO individuals view their own group members positively, they are more likely to value the norms and biases of their in‐group (Dambrun et al., 2002; Mulla et al., 2019). In their study of intimate partner violence (IPV), Mulla et al. (2019) demonstrated that for those high in SDO, group norms about the prevalence and acceptance of IPV fostered greater personal acceptance of these IPV norms. Thus, we predict that the stronger the SDO, the more compelling it is to adhere to in‐group norms, including risk information avoidance norms, thus:

Social dominance orientation will be positively related to subjective risk information avoidance norms.

Furthermore, given the emergence of anti‐science attitudes and polarized ideology within the United States during the pandemic (Jamieson & Albarracin, 2020; Plohl & Musil, 2021; Romer & Jamieson, 2020), we also explore the relationship between SDO and the perceived trustworthiness of scientists. Although SDO has never been studied with trustworthiness of scientists, its impact on other science‐related topics can offer some insight. Labota et al. (2020) found that people who are high in SDO are more willing to share conspiracy‐themed and miscellaneous culturally salient misinformation pertaining to COVID‐19. They argue that the political conservatism embedded in SDO positively associates with conspiracist ideation. People high in SDO are more likely to believe COVID‐19 is a conspiracy and are also more skeptical about science in general. This is because “conspiracies inherently entail certain groups vying for advantages over other groups” (Lobato et al., 2020, pp. 26). We suspect this conspiracy belief of science will also be appropriated onto the credibility evaluation of scientists as the scientific information source. Therefore, we predict that:

As an individual's SDO level increases, the perceived trustworthiness of scientists decreases.

The extant research on SDO also suggests that the orientation is rooted in a “competitive jungle” worldview, which maintains that people need to compete for limited resources for their survival (Federico et al., 2009). It is a social Darwinist perspective and people who have stronger SDOs are likely to support intergroup hierarchies and hold beliefs that legitimize social inequality (Perry et al., 2013). Thus, when evaluating pandemic risks, even if they recognize the risks of COVID‐19, people with a higher SDO believe in the legitimacy of intergroup inequity and may identify themselves as belonging to the “superior” group that are physically stronger, healthier, younger, and have greater access to medical care (compared to high COVID‐19 vulnerable populations who are elderly, maintain certain health conditions, or belong to systemic health and social inequity groups) and thus be less willing to adopt the precautionary behaviors needed to protect themselves as well as at‐risk groups. In such a circumstance, additional COVID‐related information might bring in cognitive dissonance, particularly if that information talks about the injustice of inequities, the need to protect others or the need to make sacrifices for public health over individual freedom. Therefore, high SDO individuals are more likely to develop attitudes that favor avoidance and hold reinforcing norms from peers residing from the same “superior” social group. Thus, we hypothesize that:

An individual's SDO will be negatively related to risk perceptions.

An individual's SDO will be positively related to attitudes toward avoidance.

Social norms

Social norms are socially constructed codes of conduct that guide, regulate, proscribe, and prescribe behaviors in particular contexts (Hechter & Opp, 2001; Rimal & Lapinski, 2015). Norms have been identified as an influential predictor of information behaviors in prior research based on the RISP and PRISM frameworks (Griffin et al., 1999; Kahlor, 2010). Previous studies categorized social norms into two general categories: injunctive and descriptive norms (Cialdini, 2007). Injunctive norms represent perceptions of social approval or disapproval of behaviors related to people's motivations for affiliations with others (Cialdini, 2007; Cialdini & Goldstein, 2004; Lapinski & Rimal, 2005). Descriptive norms are motivated by an individual's desire to do the right thing, perceiving the frequency of other's behaviors in a particular social context. Researchers interpret descriptive norms as contextual, while interpreting injunctive norms as more general (Goldstein & Cialdini, 2007). Taken together, social norms play an important role in encouraging or discouraging behaviors within a specific social context, such as risk information avoidance during COVID‐19. Similar relations have been reported by Kahlor et al. (2020), stating that people's perceived avoidance norms for carbon capture and storage information have been positively related to their risk information avoidance intentions related to the topic. Hence, the following is postulated:

Subjective risk information avoidance norms will be positively related to risk information avoidance.

METHODS

Procedure

Data were collected in the United States early in the pandemic, in April 2020 via a Qualtrics online survey panel using quota sampling on demographics to match census data and to increase the study's generalizability. Seven hundred nineteen respondents were recruited. We used two attention checking questions to filter out unattended answers and removed incomplete responses. After data cleaning, 523 valid cases were retained (see Table 1 for demographic information). The full survey took about 19 minutes to fill out. This study is based on a subset of the questions featured in the survey.

TABLE 1

Demographic information

Demographics		Percentage
Age	Age range: 18–92 years old
	Mean = 46.6 years old
Gender	Male	49%
	Female	51%
Race	Non‐Hispanic White	62%
	African American	12%
	Hispanic	17%
	Asian	5%
	American Indian or Alaskan Native	Less than 1%
	Native Hawaiian or Pacific Islander	Less than 1%
	Other	2%
Education	Less than high school degree	3%
	High school graduate (high school diploma or equivalent including GED)	19%
	Some college or an associate's (2–year) degree	30%
	Bachelor's degree (4–year)	29%
	Graduate degree	19%

The data were analyzed with Mplus7 to conduct structural equation modeling (SEM). A maximum likelihood robust estimator was used to account for issues with multivariate normality. Latent variables were constructed for SDO, attitudes toward avoidance, perceived trustworthiness of scientists, subjective information avoidance norms, affective risk response and risk information avoidance. Two‐step modeling was used to first verify the measurement model and then test the structural model. The result shows that all standardized factor loadings in the measurement model were greater than or equal to 0.74. Model fit indices include Chi‐square, comparative fit index (CFI and TLI; values close to or greater than 0.95), root mean square error approximation (RMSEA; values lower than 0.08), and standardized root mean square residual (SRMR; value lower than 0.08); which were examined as the indicators of model fit (Browne & Cudeck, 1993; Hu & Bentler, 1999). pwrSEM was used for power analysis to detect the target effects within the model (Wang & Rhemtulla, 2021) and effect sizes were evaluated according to Cohen's standards (Cohen, 1988; small = 0.2, medium = 0.5, large = 0.8).

Measures

The following variables are measured in the current study (see detailed item descriptions in Table A1). At the time the survey was fielded, the term COVID‐19 was not yet in regular use, so we used the term “the coronavirus,” as this was the wording used most often in the media at that time (The New York Times, 2020). Average variance extracted (AVE) and composite reliability (CR) were examined for discriminant validity assessment (see Table A2; AVE > = 0.5; CR > = 0.7; Fornell & Larcker, 1981).

TABLE A1

Scales information

Construct	Item	Mean	SD	Factor loading
Attitudes toward avoidance	Mark your level of agreement (Scale 1–5, from strongly disagree to strongly agree).
	Avoiding information about the coronavirus is helpful for me.	2.33	1.39	0.90
	For me, avoiding information about the coronavirus is good.	2.30	1.34	0.90
	Avoiding information about the coronavirus is beneficial for me.	2.31	1.37	0.92
	Avoiding information about the coronavirus is valuable for me.	2.28	1.36	0.93
Risk perception	Mark your level of agreement (Scale 1–5, from strongly disagree to strongly agree).
	This coronavirus is very dangerous.	4.38	0.98	–
	There is a high chance of me getting this coronavirus.	3.16	1.29	–
Affective risk response	The following statements describe my feelings about coronavirus. When I think about the coronavirus, I get… (Scale 1–5, from strongly agree to strongly disagree, reverse coded for data analysis)
	Frightened	3.43	1.35	0.84
	Tense	3.48	1.27	0.90
	Nervous	3.54	1.28	0.91
	Anxious	3.50	1.30	0.89
	Uncomfortable	3.41	1.31	0.82
Untrustworthiness of scientists	The U.S. scientists who study public health, viruses and other areas related to the coronavirus are… (Scale 1–5, from strongly disagree to strongly agree)
	Incompetent	2.44	1.34	0.80
	Poorly educated	2.19	1.28	0.86
	Unprofessional	2.20	1.27	0.85
	Inexperienced	2.30	1.34	0.86
	Hindering	2.54	1.22	0.83
	Insincere	2.45	1.27	0.89
	Unethical	2.50	1.28	0.89
	Irresponsible	2.42	1.28	0.90
	Inconsiderate	2.43	1.31	0.92
Social dominance orientation	How positive or negative do you feel about the following statements? (Scale 1–5, from very negative to very positive)
	Some groups of people are simply inferior to other groups.	2.23	1.38	0.78
	It's OK if some groups have more of a chance in life than others.	2.27	1.34	0.76
	To get ahead in life, it is sometimes necessary to step on other groups.	2.16	1.34	0.82
	Inferior groups should stay in their place.	2.15	1.38	0.84
Information avoidance norms	Indicate your level of agreement (Scale 1–5, from strongly disagree to strongly agree).
	It is expected of me that I avoid information about the outbreak.	2.25	1.39	0.89
	Most people who are important to me think that I should avoid information about the outbreak.	2.25	1.37	0.92
	Others expect me to avoid information about the outbreak.	2.24	1.36	0.93
	My family expects me to avoid information about the outbreak.	2.17	1.35	0.92
Risk information avoidance behavior	Rate your agreement or disagreement with the following statements about the coronavirus (Scale 1–5, from strongly disagree to strongly agree).
	When I hear someone talking about the outbreak, I avoid listening.	2.28	1.37	0.87
	When I see social media posts about the coronavirus outbreak, I do not read them.	2.62	1.43	0.74
	When I see news about the outbreak while surfing the Internet, I avoid it.	2.44	1.38	0.86
	I prefer not to think about the outbreak.	2.66	1.36	0.78
	I avoid watching TV news about the outbreak.	2.47	1.41	0.84
	I do not want any more information about the outbreak.	2.41	1.37	0.86
	I avoid reading about the outbreak on social media.	2.67	1.46	0.74
	I avoid learning about the coronavirus outbreak.	2.25	1.34	0.88

TABLE A2

Construct discriminant validity

Construct	AVE	CR
Attitudes toward avoidance	0.83	0.95
Affective risk response	0.76	0.94
Untrustworthiness of scientists	0.75	0.96
Social dominance orientation	0.64	0.88
Information avoidance norms	0.84	0.95
Risk information avoidance behavior	0.68	0.94

Attitude toward avoidance 1: Beliefs with respect to avoiding information about COVID‐19 were measured using an adapted version of Kahlor et al.’s (2020) 4‐item, 5‐point Likert scale. A sample item reads as: “Avoiding information about the coronavirus is beneficial for me” (M = 2.3, SD = 1.3, α = 0.95, AVE = 0.83, CR = 0.95).

Risk perception: Risk perception was measured using a 2‐item, 5‐point Likert scale to capture both the severity (“This coronavirus is very dangerous.” M = 4.4, SD  = 1.0) and vulnerability dimensions (“There is a high chance of me getting this coronavirus.” M = 3.2, SD  = 1.3). These two items’ product was computed entered as an observed variable in data analysis.

Affective risk response: In this study we based our measure of affective risk response on Witte's (1992) fear appeals research. The 5‐point Likert scale items include: “When I think about the coronavirus, I get…frightened,” “…tense,” “…nervous,” “…anxious,” and “…uncomfortable” and ranged from strongly agree (1) to strongly disagree (5) (M = 3.5, SD = 1.2, α = 0.94, AVE = 0.76, CR = 0.94).

Perceived (un)trustworthiness of scientists: Although our review of the literature describes the perceived trustworthiness of scientists, we ultimately chose to use a validated measurement index developed by Hendriks et al.’s (2015) that focuses on perceptions of scientists’ untrustworthiness (e.g., whether they are incompetent)2. Nine 5‐point Likert items were used to measure perceived untrustworthiness (M = 2.4, SD = 1.1, α = 0.94, AVE = 0.75, CR = 0.96).

Social Dominance Orientation: An individual's preference for group‐based inequality and dominance versus equality and inclusion was measured using an adapted version of Pratto et al.’s (2013) 4‐item, 5‐point Likert scale (M = 2.2, SD = 1.2, α = 0.88, AVE = 0.64, CR = 0.88).

Information Avoidance Norms: Perceived norms for risk information avoidance were measured using an adapted version of Kahlor et al.’s (2020) 4‐item, 5‐point Likert scale. A sample item is: “Others expect me to avoid information about the outbreak” (M = 2.2, SD = 1.3, α = 0.95, AVE = 0.84, CR = 0.95).

Risk Information Avoidance Behavior: COVID‐19‐related risk information avoidance behaviors were measured using an adapted version of Kahlor et al.’s (2020) 8‐item, 5‐point Likert scale. A sample item reads: “I do not want any more information about the outbreak” (M = 2.5, SD = 1.2, α = 0.94, AVE = 0.68, CR = 0.94).

RESULTS

The measurement model fit was good: χ 2 (480) = 971.65, RMSEA = 0.044 (90% CI [0.040, 0.054]), CFI = 0.96, TLI = 0.95, SRMR = 0.03. The paths proposed by hypotheses were then added into the model. Our results show that the proposed model fits the data well: χ 2 (517) = 1211.96, RMSEA = 0.051 (90% CI [0.047, 0.054]), CFI = 0.94, TLI = 0.94, SRMR = 0.07. Except for H3, H4, and H9, all the other hypotheses were supported by the data (p < 0.05). Attitudes toward avoidance (H1: b = 0.82, p < 0.001, d = 1.00) and subjective information avoidance norms (H11: b = 0.19, p < 0.001, d = 1.00) were found as positive predictors of risk information avoidance, while affective risk response (H3: b = −0.04, p = 0.095, d = 0.72) was not. Risk perception was found to positively relate to affective risk response (H2: b = 0.31, p < 0.001, d = 1.00). Perceived untrustworthiness of scientists was positively related to attitudes toward avoidance (H5: b = 0.52, p < 0.001, d = 0.68) and norms (H6: b = 0.49, p < 0.001, d = 0.64), but not significantly related to risk perception (H4: b = 0.06, p = 0.306, d = 0.04). SDO was positively related to norms (H7: b = 0.37, p < 0.001, d = 0.83), perceived untrustworthiness of scientists (H8: b = 0.56, p < 0.001, d = 1.00), risk perception (H9: b = 0.17, p = 0.004, d = 0.18), and attitudes toward avoidance (H10: b = 0.33, p < 0.001, d = 0.78). When inspecting the R‐square, this model accounted for 88% of the variance in COVID‐19 information avoidance behavior (see Figure 2 and Table 2 for path coefficients in detail; see correlation matrices in the Table A3 and A4).

FIGURE 2

Model coefficients Note: *** p < 0.001, ** p < 0.01, ns = not significant.

TABLE 2

Model statistics

Path	B	SE	p	95% CI (LL UL)		d
H1: Attitude ‐ RIA	0.82	0.04	0.000	0.738	0.897	1.00
H2: Risk perception ‐ Affective risk response	0.31	0.05	0.000	0.220	0.404	1.00
H3: Affect – RIA	−0.04	0.02	0.095	−0.082	0.007	0.72
H4: Untrustworthiness ‐ Risk perception	0.06	0.06	0.306	‐0.056	0.178	0.04
H5: Untrustworthiness ‐ Attitude	0.53	0.06	0.000	0.409	0.636	0.68
H6: Untrustworthiness ‐ Norm	0.49	0.06	0.000	0.380	0.606	0.64
H7: SDO ‐ Norm	0.37	0.06	0.000	0.249	0.483	0.83
H8: SDO ‐ Untrustworthiness	0.56	0.04	0.000	0.479	0.648	1.00
H9: SDO ‐ Risk perception	0.17	0.06	0.004	0.055	0.292	0.18
H10: SDO ‐ Attitude	0.33	0.06	0.000	0.203	0.446	0.78
H11: Norm ‐ RIA	0.19	0.05	0.000	0.085	0.294	1.00

TABLE A3

Correlation matrix of all measures

Correlations
	Attitude1	Attitude2	Attitude3	Attitude4	Norm1	Norm2	Norm3	Norm4	Risk Perception	Affect1	Affect2	Affect3	Affect4	Affect5	SDO1	SDO2	SDO3	SDO4	Untrustworthiness1	Untrustworthiness2	Untrustworthiness3	Untrustworthiness4	Untrustworthiness5	Untrustworthiness6	Untrustworthiness7	Untrustworthiness8	RIA1	RIA2	RIA3	RIA4	RIA5	RIA6	RIA7	RIA8
Attitude1	1
Attitude2	0.821	1
Attitude3	0.828	0.824	1
Attitude4	0.804	0.835	0.859	1
Norm1	0.711	0.689	0.692	0.760	1
Norm2	0.695	0.683	0.711	0.749	0.832	1
Norm3	0.717	0.692	0.708	0.729	0.810	0.869	1
Norm4	0.748	0.748	0.731	0.768	0.815	0.818	0.861	1
Risk Perception	0.191	0.180	0.167	0.175	0.224	0.208	0.210	0.227	1
Affect1	0.060	0.035	0.032	0.023	0.040	0.076	0.093	0.073	0.298	1
Affect2	0.032	0.003	0.051	0.016	0.019	0.038	0.067	0.054	0.282	0.794	1
Affect3	−0.005	−0.033	0.003	−0.018	−0.002	0.033	0.023	0.026	0.281	0.777	0.810	1
Affect4	−0.051	−0.060	−0.043	−0.056	−0.026	0.005	−0.002	−0.006	0.261	0.701	0.793	0.835	1
Affect5	−0.018	−0.027	0.000	−0.024	0.005	0.031	0.023	0.002	0.255	0.663	0.751	0.728	0.766	1
SDO1	0.375	0.361	0.375	0.430	0.425	0.438	0.441	0.437	0.174	0.039	0.057	−0.010	0.004	0.000	1
SDO2	0.388	0.363	0.385	0.405	0.389	0.391	0.397	0.426	0.125	0.018	0.015	−0.041	−0.027	−0.052	0.595	1
SDO3	0.475	0.471	0.470	0.501	0.476	0.485	0.473	0.518	0.182	−0.002	0.027	−0.044	−0.051	−0.067	0.626	0.651	1
SDO4	0.412	0.406	0.453	0.479	0.452	0.454	0.462	0.478	0.136	−0.006	0.030	−0.023	−0.046	−0.040	0.670	0.630	0.673	1
Untrustworthiness1	0.500	0.511	0.506	0.568	0.605	0.553	0.521	0.534	0.177	−0.048	−0.051	−0.059	−0.061	−0.035	0.401	0.317	0.419	0.431	1
Untrustworthiness2	0.546	0.567	0.585	0.609	0.607	0.577	0.526	0.570	0.163	−.010	−0.029	−0.049	−0.086	−0.034	0.424	0.319	0.412	0.469	0.718	1
Untrustworthiness3	0.521	0.530	0.510	0.557	0.556	0.502	0.496	0.555	0.076	−0.061	−0.068	−0.093	−0.131	−0.091	0.407	0.294	0.401	0.448	0.664	0.793	1
Untrustworthiness4	0.593	0.562	0.574	0.599	0.579	0.548	0.540	0.579	0.130	−0.019	−0.067	−0.070	−0.136	−0.068	0.392	0.360	0.442	0.466	0.666	0.785	0.788	1
Untrustworthiness5	0.559	0.558	0.596	0.597	0.594	0.571	0.578	0.573	0.123	−0.038	−0.030	−0.054	−0.083	−0.040	0.400	0.301	0.399	0.440	0.735	0.748	0.733	0.738	1
Untrustworthiness6	0.555	0.536	0.562	0.565	0.550	0.553	0.535	0.559	0.129	−0.020	−0.052	−0.060	−0.131	−0.071	0.406	0.323	0.414	0.454	0.701	0.733	0.742	0.761	0.803	1
Untrustworthiness7	0.537	0.547	0.545	0.571	0.554	0.526	0.547	0.563	0.134	−0.040	−0.055	−0.070	−0.119	−0.053	0.365	0.295	0.391	0.416	0.691	0.743	0.760	0.754	0.765	0.814	1
Untrustworthiness8	0.531	0.543	0.546	0.599	0.584	0.542	0.547	0.572	0.150	−0.027	−0.053	−0.069	−0.096	−0.051	0.411	0.313	0.411	0.444	0.725	0.766	0.765	0.772	0.826	0.834	0.875	1
RIA1	0.761	0.751	0.755	0.786	0.765	0.762	0.755	0.806	0.184	0.049	0.048	−0.012	−0.040	−0.014	0.403	0.364	0.483	0.454	0.543	0.568	0.532	0.581	0.569	0.540	0.531	0.547	1
RIA2	0.603	0.588	0.621	0.633	0.541	0.526	0.551	0.584	0.174	−0.011	−0.019	−0.051	−0.082	−0.044	0.287	0.324	0.349	0.356	0.382	0.442	0.403	0.464	0.421	0.384	0.401	0.408	0.683	1
RIA3	0.728	0.726	0.724	0.747	0.664	0.634	0.641	0.676	0.169	−0.022	−0.036	−0.063	−0.102	−0.042	0.362	0.387	0.423	0.432	0.512	0.523	0.476	0.538	0.538	0.502	0.498	0.515	0.745	0.729	1
RIA4	0.650	0.647	0.654	0.670	0.571	0.565	0.577	0.603	0.183	−0.002	0.018	−0.001	−0.031	−0.002	0.273	0.305	0.379	0.343	0.407	0.406	0.379	0.453	0.459	0.434	0.426	0.446	0.663	0.531	0.679	1
RIA5	0.691	0.711	0.718	0.730	0.612	0.593	0.602	0.618	0.198	−0.029	0.002	−0.044	−0.037	−0.004	0.286	0.291	0.370	0.373	0.481	0.472	0.409	0.457	0.499	0.447	0.445	0.463	0.707	0.583	0.736	0.721	1
RIA6	0.729	0.736	0.726	0.780	0.667	0.673	0.632	0.666	0.128	−0.046	−0.027	−0.059	−0.086	−0.037	0.384	0.350	0.447	0.434	0.505	0.527	0.487	0.501	0.539	0.509	0.504	0.495	0.735	0.603	0.715	0.693	0.753	1
RIA7	0.638	0.591	0.630	0.601	0.496	0.443	0.492	0.489	0.135	−0.045	−0.052	−0.104	−0.101	−0.056	0.218	0.261	0.292	0.269	0.365	0.407	0.373	0.423	0.432	0.358	0.359	0.389	0.566	0.697	0.703	0.598	0.633	0.640	1
RIA8	0.785	0.765	0.773	0.784	0.677	0.673	0.677	0.724	0.207	0.026	0.034	−0.019	−0.058	−0.031	0.389	0.377	0.470	0.416	0.524	0.548	0.499	0.535	0.570	0.538	0.502	0.519	0.754	0.602	0.726	0.681	0.745	0.777	0.650	1
M	2.3	2.3	2.3	2.3	2.3	2.3	2.2	2.2	13.9	3.4	3.5	3.5	3.5	3.4	2.2	2.3	2.2	2.1	2.4	2.2	2.2	2.3	2.4	2.5	2.4	2.4	2.3	2.6	2.4	2.7	2.5	2.4	2.7	2.2
SD	1.4	1.3	1.4	1.4	1.4	1.4	1.4	1.3	6.9	1.3	1.3	1.3	1.3	1.3	1.4	1.3	1.3	1.4	1.3	1.3	1.3	1.3	1.3	1.3	1.3	1.3	1.4	1.4	1.4	1.4	1.4	1.4	1.5	1.3

TABLE A4

Correlation matrix for composite measures

	Attitude	Norm	Risk	Affect	SDO	Untrustworthiness	RIA
Attitude	1.00
Norm	0.83	1.00
Risk	0.19	0.23	1.00
Affect	0.00	0.03	0.31	1.00
SDO	0.53	0.56	0.18	−0.01	1.00
Untrustworthiness	0.67	0.67	0.15	−0.08	0.52	1.00
RIA	0.88	0.78	0.20	−0.04	0.50	0.63	1.00
M	2.3	2.2	13.9	3.5	2.2	2.4	2.5
SD	1.3	1.3	6.9	1.2	1.2	1.2	1.2

DISCUSSION

Our goal for this study was to explore U.S. risk information avoidance behaviors during the COVID‐19 pandemic. To this end, we sought guidance from the PRIA framework, which suggests that risk information avoidance is driven by risk‐related cognitive, affective, and sociocultural factors. Given the heightened role of misinformation in the information landscape, we also proposed that social dominance orientation and the perceived trustworthiness of scientists would further drive risk information avoidance during the pandemic. Our results provided empirical support for the PRIA framework—particularly the relationships involving attitudes, social norms, risk perceptions, and affective risk response. Before we discuss our results in more detail, we want to acknowledge that the present study was conducted at an early stage of the COVID‐19 pandemic outbreak. Therefore, this study is considered an investigation aimed at understanding how individuals react to risk information when a particular threat is new and highly uncertain. However, due to the pandemic's continuing fermentation, people have begun to adapt to it and have normalized it as a part of a new societal routine. As the PRIA framework is a pre‐theoretical model, and the disease context continued to evolve long after data collection, structural equation modeling may be considered by some to be a premature analysis technique to apply to this cross‐sectional data. However, we chose this method based on prior empirical support found in the extant research for many of the relationships put forth in this model, as well as the strong theoretical support for all of the hypothesized relationships.

That said, consistent with PRIA, positive attitude toward avoidance and perceived normative pressure toward avoiding information (measured as subjective information avoidance norm) about COVID‐19 risks both were positively related to risk information avoidance. Meaning that, if individuals believed that avoiding COVID‐19 risk information was beneficial and that people in their social group agreed with the decision to avoid the information, they were highly likely to actively avoid the information. One potential reason for favoring risk information avoidance is that seeking COVID‐19 information might be evaluated as costly, particularly if the available messages are perceived as mixed and uncertain. Therefore, the higher cost is in the cognitive effort required to interpret and distinguish the difference between rumors and facts, as well as the emotional burden brought by uncertainty. At the time of the survey, April 2020, it was rapidly becoming clear that the incidence and fatality rate for COVID‐19 would climb precipitously, and there would be resource shortages to address the rise in cases (e.g., not enough ventilators and hospitals over capacity). The difficult cognitive and emotional burden of COVID‐19 risk information, coupled with the perceived low utility value of the available information (i.e., not many helpful preventive measures or new solutions available at the time) gave people the option to avoid the information as a strategy to cope with their emotions (Case et al., 2005; Sweeny et al., 2010).

Although previous risk information avoidance and information seeking studies have found attitudes and norms to predict information behaviors, norms were usually the most (or nearly the most) influential variable (e.g. Hovic et al., 2014; Kahlor, 2010; Kahlor et al., 2020). However, in the current context, inspecting the coefficients, attitude toward avoidance was the strongest predictor of avoidance. We wonder if the quarantine lifestyle played a part in this phenomenon, because it limited interpersonal interactions, which made it somewhat more difficult to observe others’ actual behaviors. Instead, individuals were primarily left with mass media representations of people's information behaviors and beliefs—and those seemed to come across as mixed. In such a circumstance, individuals who were ultimately driven to avoid COVID‐19 information were more driven by their own need to manage their cognitive needs than by their need to conform to social norms, although those norms did still matter.

As expected, social dominance orientation indirectly impacted risk information avoidance through attitude and social norms. Thus, if an individual holds SDO views, then they can have a positive attitude toward avoidance, and perceive positive social norms for avoiding COVID‐19 risk information. The positive attitudes and perceived supportive social norms toward avoiding COVID‐19 risk information could be due to an optimistic bias. People high in SDO believe in clear social hierarchies and favor conditions that keep their group at an advantage, compared with other groups. While SDO does not dictate group identification, it does influence attitudes that seek to maintain that group's dominance, such that high SDO individuals “favor hierarchy‐enhancing ideologies and policies” (Pratto et al., 1994, p. 742). Driven by the motivation to protect their group's superiority over others, they refuse information that might threaten this belief. However, this explanation may be undermined by the positive relationship between SDO and risk perception. As optimistic bias is triggered, people high in SDO should rate COVID‐19′s risk as lower, and this was not the case. Although the effect size of the relationship between SDO and risk perception is low (d = 0.18; Moore et al., 2013), we would like to provide an alternative explanation for the unexpected positive relationship, as well as SDO's effect on motivating avoidance.

The second tentative explanation, as argued in the literature review, is that people who are strong in SDO are more aware of the inequality between groups when facing the virus and therefore are better able to recognize the level of risk involved. However, they have a lower level of empathy for at‐risk groups (Pratto et al., 1994) and are less willing to change their behaviors or beliefs to offer help. In this situation, COVID‐19 risk information brings more emotional burden than utility value. These findings associated with SDO point out that public health risk communication is a complex subject. Individuals’ risk information behavior is not only affected by an individual's own self‐oriented concerns but also by their social ideologies and evaluations of others. Even though risk information avoidance is a personal decision, it is also influenced by intergroup beliefs.

Risk perception is positively related to affective risk response, but affective risk response does not significantly relate to risk information avoidance. The more dangerous people consider COVID‐19 to be, the more negative their emotional response is toward COVID‐19. However, different from our proposition, this risk perception does not pass onto risk information avoidance through the heightened affective response. The absence of a significant relationship could be due to the fact that this study was conducted at the early stage of the pandemic when people were more sensitive to COVID‐19 related information, regardless of their risk and affective evaluation of the virus. Meanwhile, early‐stage government messaging from the U.S. president compared the virus to the flu, which likely primed the public's perception of it as something familiar but not altogether threatening. Affective responses are a heuristic evaluation of accumulated experiences. Hence, if people's affective risk responses to the new virus were inherited from flu, it may have resulted in a disconnection between the emotion and the risk responding behavior to COVID‐19. Fear appeal research indicates that fear‐related emotions, such as worry, anxiety, and fright, can be better motivators for risk‐averse behaviors, such as information seeking, than emotion‐coping strategies, such as risk information avoidance (Slovic et al., 2004; Witte, 1992). Previous empirical data have also yielded negative correlations between fear‐related emotions and avoidance intent (Kahlor et al., 2020; Yang & Kahlor, 2013). The divergence between PRIA's proposition, previous findings, and the result from the current study makes this relationship between negative emotion and risk information avoidance more interesting to investigate. As reviewed above, response efficacy should be considered to provide richer information behind this relation.

Moving on to the perceived trustworthiness of scientists, our data show that low perceived trustworthiness (measured as untrustworthiness) impacts risk information avoidance through heightened avoidance‐related attitude and social norm evaluation. Meaning that, if individuals do not consider scientists as trustworthy, they are likely to generate a favorable attitude toward avoidance and perceive the social norms toward avoidance as more supportive and, therefore, choose to avoid the information. In epistemology literature, trusting others is argued to be a primary source in knowledge acquisition (Capet & Delavallade, 2014). Individuals are unlikely to verify every piece of the information that they encounter, making an information source's credibility a heuristic cue to be used to evaluate the quality of presented information (Hendriks et al., 2015). If this trust toward scientists is disdained and the individual believes they cannot get reliable information from a scientific source, they will need to find alternative sources and expend a high level of cognitive effort when seeking out information, or simply give up and engage in less effortful herd behavior in response to the presented risk (Smith, 2006) to justify their choice and base their decision on the behaviors of others. This mistrust also devalues the information given by scientists, resulting in a rational avoidance decision. Therefore, the supportive avoidance norm becomes salient to the individual.

Different from the hypothesis, perceived trustworthiness of scientists does not significantly relate to risk perception. However, higher SDO is associated with lower trustworthiness evaluations, indicating that people who are high in SDO are more skeptical about scientists and the evaluation of scientists’ credibility is subjected to intergroup ideology. As mentioned above in the literature review, we suspect this relationship is explained by the conspiracy belief associated with the high SDO, which affects people's benevolence evaluation of scientists and therefore undermines scientists’ credibility (Hendriks et al., 2015). Collaboratively, SDO and perceived trustworthiness of scientists explained a moderate level of variance in attitudes toward avoidance (R 2 = 0.57) and norms (R 2 = 0.58). Although SDO did contribute to risk perception, the effect size is low (Moore et al., 2013). In another words, both SDO and scientists’ trustworthiness do not affect people's evaluation of the risk event itself much but affect the evaluation of responding behaviors, which is risk information avoidance. Inspecting the determinants of coefficients, perceived trustworthiness of scientists contributes more to the attitudinal and social norm evaluations of COVID‐19 risk information avoidance.

These findings call out the importance of trust in risk communication—it can affect how people perceive risk and to what degree they are willing to accept the message delivered from a particular source (Liao et al., 2018; Slovic, 1997). As put by Slovic (1997), “In the absence of trust, science [and risk assessment] can only feed public concerns by uncovering more bad news.” Researchers termed the trustworthiness that people hold toward experts as “epistemic trustworthiness” and argue that it is an important pillar in information exchange (Capet & Delavallade, 2014). However, trust is harder to gain than to lose as it is built by a long‐time accumulation of doing the right thing, while having the potential to collapse after just a few instances of misconduct (Slovic, 1993).

An epistemic trustworthiness study reports that three important dimensions for laypeople when judging the credibility of experts are practical intelligence, virtuous character, and good will (Rapp, 2011). Risk communication practitioners can consult on these three dimensions for inspiration when constructing and maintaining the credibility of information sources. Except for the long‐term efforts to build a golden reputation, it is also important for communication practitioners and media to protect and even establish scientists’ credibility during a crisis to build reliable public figures as information sources to foster effective risk communication. Moreover, as mentioned earlier, mistrust of scientists not only impacts how people evaluate the particular risk but also affects their judgment of solutions, such as vaccination. Progress is being made in vaccine development since the outbreak. Once the COVID‐19 vaccine was available, the baton from bioscience was passed to social science to promote vaccination. Lessons from previous influenza pandemics have shown how distrust in authorities can act as a barrier to getting individuals on board with vaccination protocols (Schmid et al., 2017), and this has indeed been the case.

LIMITATIONS AND FUTURE RECOMMENDATIONS

As with all studies, this study has a few limitations. Despite the aforementioned limitations with the study timing (early in the pandemic) and research analysis/method, we also note the following limitations: First, our data show SDO is significantly related to risk perception, but the power is small. Future research is needed to test this relationship, probably in a different context other than COVID‐19, to develop a solid explanation for it. Second, our information avoidance construct may conflate a variety of factors, such as inertia, intention, and learning under information avoidance. Future survey efforts could use or be informed by a validated information avoidance scale, like the one from Howell and Shepperd (2013). Third, as mentioned in the discussion section, the fear appeal literature suggests that efficacy could play an important role in explaining the relationship between affect and avoidance. Therefore, we recommend future studies to take efficacy into consideration.

In addition, this study only examined injunctive and descriptive norms mentioned in the PRIA. Additional research will be needed to further understand the “jingle‐jangle fallacies” (Deline & Kahlor, 2019, p8) mentioned in the PRIA about social norms.

CONFLICT OF INTERESTS

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.