Experience,experts,statistics,or just science? Predictors and consequences of reliance on different evidence types during the COVID-19 infodemic

Abstract

As an unprecedented global disease outbreak, the COVID-19 pandemic is also accompanied by an infodemic. To better cope with the pandemic, laypeople need to process information in ways that help guide informed judgments and decisions. Such information processing likely involves the reliance on various evidence types. Extending the Risk Information Seeking and Processing model via a two-wave survey (N = 1284), we examined the predictors and consequences of US-dwelling Chinese’s reliance on four evidence types (i.e. scientific, statistical, experiential, and expert) regarding COVID-19 information. Overall, Risk Information Seeking and Processing variables such as information insufficiency and perceived information gathering capacity predicted the use of all four evidence types. However, other Risk Information Seeking and Processing variables (e.g. informational subjective norms) did not emerge as important predictors. In addition, different evidence types had different associations with subsequent disease prevention behaviors and satisfaction with the US government’s action to address the pandemic. Finally, discrete emotions varied in their influences on the use of evidence types, behaviors, and satisfaction. The findings provide potentially valuable contributions to science and health communication theory and practice.

Keywords

COVID-19 emotion evidence information processing

1. Introduction

People rely on different information to arrive at judgments about risks. Such information often includes various types of evidence, which may demand different amounts of effort in information processing to form judgments. Basing risk judgments on one’s own experience may require little energy or resource, whereas understanding complex statistical information will likely call for more exertion. It is important to note that the upstream predictors and downstream consequences of the use of evidence types in science and risk communication are not well understood, although different types (e.g. scientific, statistical, experiential, expert) are frequently employed in individuals’ information processing in these contexts and may produce differential impacts on attitudes and behaviors (Hornikx, 2005).

Notably, the unprecedented COVID-19 pandemic is also accompanied by an infodemic, which is worsened by rumors and misinformation spreading rapidly on social media. As a result, people frequently encounter COVID-19 information that draws on different evidence types, which may or may not help them make informed decisions. In addition, due to much uncertainty surrounding the disease and the rapidly changing nature of the pandemic situation, even information coming from legitimate sources can become outdated quickly, making it insufficient to use the credibility heuristics for information seeking and processing. Because pandemic-related information is critically important to one’s health, life or death in some situations, properly evaluating and processing such information, including its use of evidence, seem more crucial than ever.

We focus on the information processing of a special population (i.e. US-dwelling Chinese) during the COVID-19 pandemic for this investigation. US-dwelling Chinese, including Chinese immigrants and Chinese nationals living in the United States, have experienced increased verbal assaults, aggression, and hate crimes against them since the start of the COVID-19 pandemic (Asian Pacific Policy and Planning Council, 2020). As one of the fastest growing ethnic groups in the United States, US-dwelling Chinese are especially vulnerable during the pandemic because they have to deal with a double threat, the virus and the discrimination. How this group processes pandemic-related information, specifically how they use different evidence types for judgments and decision-making, is particularly important for the development of their pandemic-related attitudes and the adoption of disease prevention behaviors.

Drawing on the Risk Information Seeking and Processing (RISP) model (Griffin et al., 1999) as the primary theoretical framework, we attempt to achieve two primary objectives. The first is to examine the various predictors of US-dwelling Chinese’s reliance on four evidence types (i.e. scientific, statistical, experiential, and expert) of COVID-19 information. Previous information processing research, including RISP studies, has predominantly adopted the heuristic-systematic model (HSM; Eagly and Chaiken, 1993), which regards information processing as consisting of two modes. In this study, we further differentiate the evidence types that people usually use when processing science information. This differentiation is important because both modes of processing are a continuum and can vary in their own extensiveness (Chaiken et al., 1989). That is, not all systematic or heuristic processing is the same; some may use more effort than others. By integrating persuasion literature on evidence types (Hornikx, 2005), we will be able to take a more nuanced look at information processing. In addition to investigating commonly used RISP factors as the predictors, we also consider the uniqueness of US-dwelling Chinese who have access to information channels from more than one country. Thus, their beliefs in both United States and Chinese channels will be explored as competing or complementary predictors of reliance on different evidence types. Furthermore, although affective responses to the risk have been examined as discrete emotions in other RISP studies (e.g. Yang et al., 2019), we include a wider range of discrete emotions in this study and incorporate literature on the effects of discrete emotions on information processing. We consider these modifications to the RISP model as a test of its validity across contexts, conceptualization, and operationalization.

The second objective of this study is to investigate the linkage between reliance on different evidence types and pandemic-related attitudes and the adoption of disease prevention behaviors. In addition to predictors of information seeking and processing, the original RISP model (Griffin et al., 1999) also connects information seeking and especially processing to downstream variables such as attitudes and behavioral intentions adopted from the theory of planned behavior (TPB; Ajzen, 1991). This segment of the RISP model has received limited attention. In this study, we examine two downstream variables (i.e. satisfaction with the US government’s action to deal with coronavirus and adoption of disease prevention behaviors) as part of the TPB-based segment of the RISP model. We consider satisfaction as a proxy for attitudes toward actions to address coronavirus. Moreover, since our sample is US-dwelling Chinese, their satisfaction with the US government’s action will likely reflect an additional layer of attitude that may be described as intercultural. On the contrary, disease prevention behaviors are what individuals do to protect themselves from getting infected with the coronavirus, representing the ultimate behavior in the TPB. By establishing the nexus of information processing and attitudes and behaviors, we aim to address the “so what” question raised for the RISP model (Griffin et al., 2013).

To accomplish the two objectives, we conducted a two-wave survey among a sample of US -dwelling Chinese during the COVID-19 pandemic to assess their reactions to the pandemic, including the RISP predictors of information processing at Wave 1 and reliance on different evidence types, satisfaction, and behaviors at Wave 2. Notably, unlike most RISP studies that focused on behavioral intentions, the two-wave nature of this study allows us to measure actual processing and disease prevention behaviors at Wave 2. Apart from the aforementioned theoretical contributions we intend to make, our study also aims to offer practical implications on science literacy education, journalistic practice, and public health guidelines during and beyond pandemics.

2. RISP model

Drawing from well-established theoretical models, such as the heuristic-systematic model and the theory of planned behavior, the RISP model presents a series of cognitive and affective factors associated with RISP and, more importantly, delineates the connections among them (Griffin et al., 1999). Two decades of research has provided substantive support for the RISP model in a wide range of contexts (Griffin et al., 2013; Yang et al., 2014). The two major outcomes in the RISP model are information seeking and processing, both of which can occur through more or less effortful routes (Griffin et al., 1999).

We focus mainly on the information processing part of the RISP model, which is deeply rooted in the HSM. Specifically, according to the HSM, there are generally two modes of information processing (i.e. systematic vs heuristic). Systematic processing involves a comprehensive, analytic approach to accessing and inspecting information for its relevance and importance, and integrating all useful information in making judgments (Chaiken et al., 1989). This mode of information processing requires some level of mental effort and cognitive capacity. In contrast, heuristic processing is qualitatively different. While heuristic processing does not require effort or resource like systematic processing, it is not simply the opposite of systematic processing because the level of effort is less relevant (Chaiken et al., 1989). Heuristic processing is regarded as the default mode of processing in which people rely on learned knowledge structures by way of simple decision rules or cognitive heuristics to form judgments (Chaiken et al., 1989). These cognitive heuristics often include dependence on existing knowledge and experience, experts’ opinions, perceived social consensus, and length of the message. Individuals are more likely to engage in heuristic processing when they perceive low issue involvement, low processing capacity, or little need for systematic processing (Chaiken et al., 1989). Because systematic and heuristic processing belong to two different continuums, they can occur independently or co-occur in an additive or interactive manner. Notably, information processing can influence attitudes toward an issue or an object, with systematic processing resulting in a stronger and more stable attitude and heuristic processing leading to a weaker and less stable attitude (Chaiken et al., 1989).

The RISP model puts forward multiple motives that can stimulate information processing (Griffin et al., 1999). First, integrating accuracy motivation, sufficiency principle, and judgmental confidence from the HSM (Eagly and Chaiken, 1993), the RISP model describes information insufficiency as the gap between one’s current knowledge level of a risk and the information sufficiency threshold one needs to reach to feel sufficiently confident in addressing that risk (Griffin et al., 1999). Information insufficiency is proposed as a primary motivation for information processing such that the greater the insufficiency, the more likely people are to adopt systematic processing to achieve accuracy and confidence in their judgments and the less likely they are to engage in heuristic processing (Griffin et al., 1999).

Adopting concepts of perceived behavioral control and subjective norms from the TPB (Ajzen, 1991) and the idea of capacity from the HSM (Eagly and Chaiken, 1993), the RISP model also proposes perceived information gathering capacity and informational subjective norms as antecedents of information processing (Griffin et al., 1999). Perceived information gathering capacity describes one’s perceived ability to process the desired information (Griffin et al., 1999). The more capable individuals feel of processing information, the more likely they are to adopt systematic processing and the less likely to use heuristic processing (Griffin et al., 1999). The RISP model defines informational subjective norms as the perceived social pressure to know about an issue (Griffin et al., 1999). The stronger the informational subjective norms, the more likely people are to adopt systematic processing and less likely to turn to heuristic processing. Informational subjective norms can influence information processing directly or indirectly through affecting information insufficiency (Griffin et al., 2013).

Relevant channel beliefs, defined as beliefs about channels of risk information, such as trustworthiness and usefulness, could influence information processing strategies people employ (Griffin et al., 1999). Notably, relevant channel beliefs is the least studied RISP predictor of information seeking and processing, possibly due to the difficulty of finding consistent conceptualization and operationalization (Griffin et al., 2013). The few RISP studies that examined this variable have also generated inconsistent results (Griffin et al., 2013). Instead of treating information channels as one entity, we examine beliefs in relevant information channels from both the United States and China (i.e. Centers for Disease Control and Prevention (CDC), government, scientists, media) because our target population has access to both US and Chinese channels, which often provide different information regarding COVID-19 (e.g. Yuan, 2020). In this study, we focus on the trustworthiness or credibility aspect of each channel, which has been shown to influence information processing (Trumbo and McComas, 2003).

In addition, the original RISP model includes risk perception and affective responses as distal factors indirectly influencing information processing through other proximal factors (e.g. information insufficiency) (Griffin et al., 1999). Risk perception is often conceptualized as one’s perceived probability of being affected by a risk combined with the perceived severity of the risk and proposed as a precursor of affective responses to a risk (Griffin et al., 1999). As for affective responses, research has provided ample support for their direct influence on information processing (Griffin et al., 2013), which is not accounted for in the original RISP model (Griffin et al., 1999). However, no clear theoretical explanations for such a direct linkage have been provided, which is necessary before it can be formally added to the RISP model. In the following section, we will review relevant literature to propose some theoretical explanations for a direct relationship between affective responses (i.e. discrete emotions) and information processing.

With regard to empirical RISP studies investigating predictors of information processing, past research has consistently shown that information insufficiency, perceived information gathering capacity, informational subjective norms, and negative affect are positively related to systematic processing, supporting the RISP model’s propositions (e.g. Griffin et al., 2013; Yang et al., 2014). In contrast, findings on heuristic processing have been less consistent and sometimes contradictory to the RISP model’s propositions (e.g. Griffin et al., 2013; Yang et al., 2014). These less clear results might be attributed to the low reliability of different heuristic processing scales used in empirical investigations, calling for more research into the measurements of information processing (Griffin et al., 2013; Yang et al., 2014).

3. Reliance on different evidence types

Although the HSM provides definitions of systematic and heuristic processing (Eagly and Chaiken, 1993), the operationalizations of these two processing modes in surveys vary substantially across studies, with some tapping into the notion of deep versus superficial engagement with information and others assessing different focus on message content (Schemer et al., 2008). These operational differences may be partly attributed to the extensiveness of each processing mode. For instance, commonly used systematic processing scales might most likely capture the upper end of a systematic processing continuum rather than the lower end that represents “more than marginal levels of effort and cognitive capacity” (Chaiken et al., 1989: 212).

In a study that examined the relationships between information processing and risk judgment, Trumbo (1999) operationalized systematic processing as reliance on scientific and statistical information and heuristic processing as use of experts’ recommendations and one’s existing knowledge and experience. This operationalization was based on systematic processing being “most strongly characterized by greater effort and the desire to evaluate information formally” and heuristic processing being “most strongly characterized by low effort and reliance on existing knowledge” (Trumbo, 1999: 394). While this operationalization is intuitively appealing and supported by the data (Trumbo, 1999), we take the stance that any type of message content can be considered as heuristic cues. Specifically, reliance on statistical information does not necessarily suggest that individuals would process this information systematically. On the contrary, individuals could still use statistics as a heuristic cue to aid their judgments without expending much effort in evaluating the information. For example, simply following the number of COVID-19 cases in one’s state does not require exertion despite the use of statistical evidence.

Recognizing the abovementioned limitations, we believe that there is value to go beyond systematic and heuristic processing and focus on what message content individuals are likely to rely on in their information processing. While reliance on certain message content is more likely to be an indication of one information processing mode than the other, it will not be known for sure unless the way in which individuals process this content is also assessed. Whereas both systematic and heuristic processing can contribute to positive outcomes for the individual depending on an individual’s predispositions (Chaiken et al., 1989), investigating what content individuals are likely to rely on and what effects this reliance will have can provide important implications for the development of a message.

Relevant to this is the persuasion literature on the effectiveness of various evidence types in a message (Hornikx, 2005). Specifically, this literature has examined evidence types, including anecdotal, statistical, causal, and expert evidence. It was generally found that anecdotal evidence is the least persuasive type, but the other three types do not differentiate from one another in their effectiveness (Hornikx, 2005). In this study, we focus on four evidence types (i.e. scientific, statistical, expert, and experiential). Scientific evidence is defined as information generated from rigorous scientific methods (Jeong and Songer, 2008), whereas statistical evidence refers to a numerical summary of a series of instances (Rieke and Sillars, 1984). Statistical evidence generated from scientific methods can be considered as one type of scientific evidence. In comparison, expert evidence consists of an expert’s testimony (Rieke and Sillars, 1984), while experiential evidence derives from a person’s past experience and existing knowledge (Tonelli, 2006). As no new information is needed from the outside, reliance on past experience may represent the most prototypical mode of heuristic processing (Kelley, 1999). Overall, these four types exemplify a diverse range of evidence types commonly utilized when individuals process science and risk messages during and outside pandemics. Because of their differences, the reliance on each type may be associated with different predictors and consequences.

Unsurprisingly, RISP research focusing on information processing has concentrated on systematic or heuristic processing or both. The novelty of this study is to extend the applicability of the RISP model to predict the use of different evidence types in information processing. It should be noted that we do not argue that these four evidence types are necessarily exclusive to one another or that when processing a message, individuals will rely on only one type of evidence. Conceptualizing the use of four evidence types as different information processing strategies, we aim to explore their relationship with RISP variables, including informational subjective norms, information insufficiency, perceived information gathering capacity, and trust in relevant channels. For example, for the predictor—trust in relevant channels—we expect that trust in scientists and the CDC may show some association with the use of scientific and statistical evidence, whereas the relationship between trust in the government and media and use of evidence types is less clear.

In this study, we are also interested in examining the consequences of reliance on different evidence types. According to the RISP model, processing information about a risk behavior could influence the strength and stability of attitudes toward the behavior and the behavior itself (Griffin et al., 1999). Specifically, research has shown that systematic processing can also lead to more adoption of health-protective behaviors (Hovick et al., 2011). In addition, the persuasion literature has shown that the use of different evidence types could influence a variety of attitudinal and behavioral outcomes through various underlying mechanisms (e.g. statistical evidence could indicate objectivity and thus be perceived as more useful; Hornikx, 2005). In this study, we examine how reliance on different evidence types may influence satisfaction with the US government’s action to deal with coronavirus, an attitudinal outcome, and adoption of disease prevention behaviors, a behavioral outcome.

4. Discrete emotions and information processing

Besides the RISP model, there are many theories conceptualizing how emotions directly influence information processing. The Appraisal Tendency Framework (ATF) assumes that discrete emotions are differentiated by a number of appraisal dimensions and can shape the appraisal of future events in line with the appraisal dimensions underlying those emotions (Han et al., 2007). In particular, the ATF proposes that, depending on the appraisal dimensions characterizing discrete emotions, discrete emotions can influence the depth of information processing (Han et al., 2007). For instance, emotions with high-certainty appraisals, such as anger and disgust, will likely lead to heuristic processing, whereas emotions that have low-certainty appraisals, including fear and hope, are associated with systematic processing (Ferrer et al., 2016). This is because certainty results in less motivation to process information carefully or pay attention to details (Ferrer et al., 2016).

The ATF is useful as we explore a wide range of discrete emotions as predictors of reliance on evidence types in the context of COVID-19 because the emotions people feel in response to the pandemic can guide their judgments and decision-making regarding the pandemic. Fear is experienced when there is an imminent threat, disgust is felt when being too close to an indigestible object, anger is elicited when experiencing a demeaning offense, sadness responds to irrevocable loss, compassion is induced when seeing others’ suffering, elevation is caused by witnessing exemplars of moral beauty, and hope is triggered by a yearning for a better future (Haidt, 2003; Lazarus, 1991). These emotions are chosen because they are relevant to the COVID-19 pandemic. Individuals will likely experience fear because of the threat posed by COVID-19, disgust as a result of sensitivity to diseases, anger due to pandemic-related injustice, sadness and compassion resulting from witnessing COVID-19-related suffering, elevation owing to exposure to inspiring stories during the pandemic, and hope by virtue of seeing effective actions taken to address the pandemic. Notably, while most of the emotions we examine could be categorized as having high- versus medium- versus low-certainty appraisals and thus be linked with systematic or heuristic processing, we do not propose specific hypotheses regarding how they may be associated with the use of evidence types. This is in line with our conceptualization of these evidence types, all of which could indicate more or less effortful processing depending on the situation.

5. Research questions

Based on the aforementioned literature review on the RISP research, we propose the following research question instead of hypotheses due to lack of research examining the predictors of the use of the four evidence types:

RQ1. Will (a) informational subjective norms, (b) information insufficiency, (c) perceived information gathering capacity, and (d) trust in relevant channels be associated with reliance on scientific, statistical, experiential, and expert evidence types?

We also propose the following research question to investigate the potential link between the use of evidence types and attitudinal and behavioral outcomes:

RQ2. Will reliance on scientific, statistical, experiential, and expert evidence types be associated with (a) adoption of disease prevention behaviors and (b) satisfaction with the US government’s action?

In addition to information processing, much research has found an important role played by discrete emotions in influencing health-related attitudes and behaviors (Ferrer et al., 2016). To specifically explore the impacts of the abovementioned discrete emotions on these outcomes, we propose a research question below:

RQ3. Will (a) fear, (b) disgust, (c) anger, (d) sadness, (e) compassion, (f) elevation, and (g) hope be associated with reliance on scientific, statistical, experiential, and expert evidence types, adoption of disease prevention behaviors, and satisfaction with the US government’s action?

Figure 1 presents the theoretical framework investigated in this study.

Figure 1.

Theoretical framework.

6. Method

Sample and procedure

The data for this study originated from a two-wave online survey assessing US-dwelling Chinese’s experiences and opinions concerning COVID-19. The two-wave survey conducted between 16 March 2020 and 16 April 2020 was fielded to a convenience sample of US-dwelling Chinese, who were recruited by the authors via online discussion boards, social media (e.g. WeChat), and emails. Those who were Chinese people living in the United States at Wave 1 were eligible to participate. Our recruitment of participants for the first wave survey lasted for about 2 weeks, meaning that many participants answered the first survey on different dates. Participants who finished the first wave survey were invited via emails about 1 week after completing the first survey. Only participants who provided emails and consent to further correspondence at Wave 1 were invited for Wave 2. Incentives for participation were provided: 16 participants at Wave 1 and 10 at Wave 2 were randomly selected to receive a US$100 Amazon gift card. The questionnaires were delivered in Chinese. Survey measures adopted from existing research reported in English (described below) were translated by the lead author and validated with back-translation by the other authors, who are proficient in both English and Chinese.

A total of 2208 participants completed the survey at Wave 1 and passed two attention check questions (e.g. “Please select the disagree option”). A total of 1295 participants completed the survey at Wave 2¹ and passed one attention check question. Responses from 11 participants were dropped as they were not in the United States at Wave 2, resulting in a final sample of 1284 participants. Among these participants, 51.2% identified as male (n = 658), 48.1% as female (n = 618), and the rest as other genders or did not respond to the question (n = 8). The average age was 26 years old (SD = 4.36). Most participants had a bachelor’s degree or higher (92.6%), and their median annual household income was between US$25,000 and US$34,999. Only 10.5% (n = 135) were US citizens or permanent residents. Participants on average had lived in the United States for 4.71 years (SD = 4.02).

Measures

Table 1 presents the descriptive statistics and composite reliabilities of our measures. Wave 1 measures included risk severity and susceptibility (Hovick et al., 2011), fear, disgust, anger, sadness, compassion, elevation, hope, current knowledge (Griffin et al., 2008), information sufficiency threshold (Lu et al., 2020), informational subjective norms (Griffin et al., 2008), trust in relevant channels (Harring, 2013), perceived information gathering capacity (Griffin et al., 2008), and education. Wave 2 measures included reliance on four evidence types (i.e. scientific, statistical, experiential, and expert) (Trumbo, 1999), disease prevention behavior, and satisfaction with the US government’s action. We included education as a control variable for reliance on four evidence types because as an indicator of one’s existing knowledge structures, education has been shown to influence information processing (Kahlor et al., 2006).

Table 1.

Descriptive statistics of survey measures.

Latent variable
Observed indicators	M	SD
Perceived information gathering capacity
If I wanted to get more information about coronavirus, I could readily take the time to gather any additional information I might need. (1 = disagree, 5 = agree)	4.02	.87
. . . I would know how to separate facts from rumors	3.89	.91
. . . I would know where to go for more information.	3.97	.88
Observed variable	M	SD
Risk perception. This was calculated based on the multiplication of perceived severity and perceived susceptibility (detailed items listed below)	9.86	3.96
Perceived severity. How severe do you think the consequence of getting infected with coronavirus is? (1 = not severe at all, 5 = extremely severe)	3.92	.85
Perceived susceptibility. How likely do you think you are to get infected with coronavirus? (1 = extremely unlikely, 5 = extremely likely)	2.50	.81
When it comes to the coronavirus pandemic in the past week, how much do you feel . . . (1 = not at all, 5 = a great deal)Scared?	3.11	.94
Disgusted?	2.47	1.26
Angry?	3.17	1.14
Sad?	3.34	1.07
Compassionate?	3.81	.92
Elevated?	1.84	.90
Hopeful?	3.27	.98
Informational subjective norms. People who are important to me (e.g. parents, friends, etc.) expect me to know something about coronavirus. (1 = disagree, 5 = agree)	4.37	.74
Current knowledge. How much do you think you currently know about coronavirus? (1 = not at all, 5 = a great deal)	3.46	.67
Information insufficiency. How much more, if at all, do you think you need to know about coronavirus in order to deal with it adequately? (1 = no need to know more at all, 5 = need to know a lot more)	3.56	.97
Trust in relevant channels. How much trust do you have in the US Centers for Disease Control and Prevention to give you accurate information about coronavirus? (1 = no trust at all, 5 = complete trust)	3.20	.88
. . . US government . . .	2.65	.92
. . . US scientists . . .	3.75	.72
. . . US media . . .	2.81	.86
. . . Chinese Centers for Disease Control and Prevention . . .	3.46	.89
. . . Chinese government . . .	3.26	.97
. . . Chinese scientists . . .	3.78	.75
. . . Chinese media . . .	2.97	.90
Education. What is the highest education level you have earned? (1 = elementary school and below, 2 = middle school or high school, 3 = bachelor’s degree, 4 = master’s degree, and 5 = doctorate degree (e.g. PhD, JD, and MD))	3.69	.83
Scientific type. In the past week, how often have you made a judgment on coronavirus based on scientific information with a great deal of detail? (1 = not at all, 5 = all the time)	3.51	1.04
Statistical type. In the past week, how often have you made a judgment on coronavirus based on statistical information?	3.87	.96
Observed variable	M	SD
Experiential type. In the past week, how often have you made a decision about coronavirus based on your previously existing knowledge and experience?	3.90	.94
Expert type. In the past week, how often have you placed your trust in the experts and gone with their recommendation on coronavirus?	3.80	.90
Disease prevention behaviors. In the past week, how often have you taken action to protect yourself from getting infected with coronavirus?	4.65	.66
Satisfaction with the US government’s action. How satisfied are you with the US government’s action to deal with coronavirus in the past week? (1 = unsatisfied, 5 = satisfied)	2.39	.99

SD: standard deviation.

Analyses

We employed structural equation modeling (SEM) in Mplus 8.3 with latent and observed variables to investigate the proposed research questions and assess the overall model fit. We followed a two-step modeling procedure by first constructing a measurement model and then a structural model (Kline, 2016). We modified the structural model for model fit improvement by following suggestions from modification indices and considering theoretical reasoning simultaneously.

Before conducting SEM, we screened data for collinearity, normality, and missing values. Since all tolerance values were above .10 and variance inflation factor (VIF) values were below 10, collinearity was not a concern (Kline, 2016). We used Stata 16 to test for multivariate normality and found possible violations of the multivariate normality assumption (i.e. skewness > |3| and kurtosis > |20|; Kline, 2016). Therefore, we employed a maximum likelihood estimator with robust standard error (MLR), which also handled missing values, for its robustness to non-normality.

7. Results

Model fitting results

Measurement model

We created a measurement model of one latent variable: perceived information gathering capacity. Since this model was just identified, no model fit was generated. We moved next to the structural model.

Structural model

We built the structural model based on the proposed relations in Figure 1. The initial structural model did not fit the data well: χ²(222) = 2899.95, p < .001, root mean square error of approximation (RMSEA) = .097 (90% confidence interval (CI) = .094–.100), comparative fit index (CFI) = .464, standardized root mean square residual (SRMR) = .082. In light of modification indices, theoretical reasoning, and empirical evidence from previous research, we refined the initial structural model in the following ways. First, we added 14 sets of covariance to the error terms of different discrete emotion items to account for their common method variance and intrinsic association with arousal. Second, we covaried all the error terms of the four reliance on evidence type items because they all represented information processing. Third, we covaried all the error terms of the three perceived information gathering capacity items because they were the indicators of the same latent variable and highly correlated. Fourth, we added a path from risk perception to information insufficiency. This path indicates that rather than working through affective response, risk perception can have a direct relationship with information insufficiency (Lu et al., 2020). Fifth, we added paths from the eight trust in relevant channels variables and information insufficiency to both disease prevention behavior and satisfaction. Channels beliefs, including perceived trustworthiness and usefulness, have been shown to influence health-related attitudes and behavioral intentions (Yang et al., 2010). It is likely that information from these channels helps shape individuals’ attitudes and intentions, and the amount of information obtained partly depends on channel beliefs (Yang et al., 2010). As for information insufficiency, it is possible that this subjective “gap” in knowledge directly influenced attitudes and behaviors because it represents some level of uncertainty that served as a behavioral motivator (Hogg, 2000). Finally, we added paths from education and information subjective norms to disease prevention behavior. As an indicator of one’s existing knowledge structure, it is no surprise that education is associated with the adoption of disease prevention behaviors (Hovick et al., 2011). Informational subjective norms were associated with behavior probably because it served as a substitute for subjective norms in the model (Ajzen, 1991).

The final structural model fit the data adequately: χ²(178) = 472.75, p < .001, RMSEA = .036 (90% CI = .032–.040), CFI = .941, SRMR = .041. Overall, this final model explained 23.2% of the variance in reliance on scientific type, 14.2% in statistical type, 14.4% in experiential type, 10.6% in expert type, 15.1% in adoption of disease prevention behavior, and 16.9% in satisfaction with US government’s action.

Relation testing results

We report unstandardized coefficients with standard errors for all relationships investigated in the final structural model in Figure 2 and Tables 2 and 3.

RQ1(a)–(d) asked about the relationships between RISP variables and reliance on the four evidence types. We found that information insufficiency and perceived information gathering capacity were positively associated with the use of each of the four evidence types. In addition, trust in US media was negatively related to the use of statistical evidence, trust in Chinese scientists was positively related to experiential and expert evidence, and trust in US scientists was also positively related to expert evidence.

RQ2(a)–(b) centered around the relationships between reliance on evidence types and behavior and satisfaction. The results show that the reliance on experiential and expert evidence was positively related to the adoption of disease prevention behaviors, whereas the use of scientific evidence was negatively associated with satisfaction.

RQ3(a)–(g) concerned the relationships between discrete emotions and the outcome variables, including reliance on evidence types, behaviors, and satisfaction. We found that the reliance on scientific evidence was positively related to disgust and compassion, experiential evidence was positively related to anger and compassion, and expert evidence was positively related to hope. Furthermore, the adoption of disease prevention behaviors was positively related to fear and compassion, but negatively to sadness. As for satisfaction, it was negatively associated with anger and positively associated with elevation and hope.

Figure 2.

Unstandardized estimates and standard errors (in parentheses) for the final structural model.

Table 2.

Path estimates predicting reliance on evidence types (B (SE)).

	Scientific type	Statistical type	Experiential type	Expert type
Trust in the US CDC	−0.03 (0.04)	−0.02 (0.04)	−0.01 (0.04)	0.02 (0.04)
Trust in the US government	−0.02 (0.04)	0.07 (0.04)	0.04 (0.04)	0.00 (0.04)
Trust in the US scientists	0.03 (0.05)	0.08 (0.05)	−0.06 (0.05)	0.15** (0.04)
Trust in the US media	−0.02 (0.04)	−0.09* (0.04)	−0.06 (0.04)	0.05 (0.04)
Trust in the Chinese CDC	−0.06 (0.05)	−0.04 (0.05)	−0.07 (0.04)	0.01 (0.05)
Trust in the Chinese government	0.03 (0.05)	−0.01 (0.05)	0.02 (0.04)	0.02 (0.05)
Trust in the Chinese scientists	0.01 (0.05)	0.04 (0.05)	0.12* (0.05)	0.12* (0.05)
Trust in the Chinese media	−0.01 (0.05)	0.04 (0.05)	0.01 (0.04)	0.01 (0.04)
Perceived information gathering capacity	0.95*** (0.15)	0.65*** (0.11)	0.62*** (0.1)	0.31*** (0.08)
Information insufficiency	0.17*** (0.03)	0.12*** (0.03)	0.07* (0.03)	0.09** (0.03)
Informational subjective norms	0.04 (0.04)	0 (0.04)	0.07 (0.04)	0.06 (0.04)
Fear	−0.01 (0.03)	0.03 (0.03)	−0.02 (0.03)	0.00 (0.03)
Disgust	0.06* (0.03)	0.03 (0.02)	0.01 (0.02)	0.01 (0.02)
Anger	0.05 (0.03)	0.04 (0.03)	0.06* (0.03)	0.03 (0.03)
Sadness	−0.01 (0.03)	0.02 (0.03)	−0.03 (0.03)	0.02 (0.03)
Compassion	0.1** (0.04)	0.06 (0.04)	0.13*** (0.04)	0.04 (0.03)
Elevation	0.05 (0.03)	0 (0.03)	−0.02 (0.03)	0.02 (0.03)
Hope	−0.03 (0.03)	−0.03 (0.03)	0.01 (0.03)	0.08** (0.03)
Education	0.04 (0.03)	0.03 (0.03)	0.08** (0.03)	0.01 (0.03)

CDC: Centers for Disease Control and Prevention.

p < .05; **p < .01; ***p < .001.

Table 3.

Path estimates predicting satisfaction with the US government’s action and disease prevention behaviors (B (SE)).

	Satisfaction with the US government’s action	Disease prevention behaviors
Trust in the US CDC	0 (0.04)	−0.05* (0.03)
Trust in the US government	0.2*** (0.04)	−0.02 (0.03)
Trust in the US scientists	0.04 (0.05)	−0.08** (0.03)
Trust in the US media	0.09* (0.04)	0.03 (0.03)
Trust in the Chinese CDC	−0.03 (0.05)	0.01 (0.03)
Trust in the Chinese government	−0.07 (0.05)	0.03 (0.03)
Trust in the Chinese scientists	−0.02 (0.05)	0.03 (0.04)
Trust in the Chinese media	−0.13** (0.04)	0.01 (0.03)
Scientific type	−0.09** (0.03)	−0.01 (0.02)
Statistical type	0 (.04)	0.02 (0.03)
Experiential type	−0.03 (0.03)	0.13*** (0.02)
Expert type	0.05 (0.03)	0.07** (0.02)
Fear	−0.03 (0.03)	0.11*** (0.02)
Disgust	0 (0.03)	−0.01 (0.02)
Anger	−0.08** (0.03)	−0.01 (0.02)
Sadness	−0.01 (0.03)	−0.05* (0.02)
Compassion	0.03 (0.03)	0.05* (0.02)
Elevation	0.07* (0.03)	−0.03 (0.02)
Hope	0.08** (0.03)	−0.03 (0.02)
Information insufficiency	−0.09** (0.03)	0.04* (0.02)
Informational subjective norms	−	0.08** (0.02)
Education	−	0.08*** (0.02)

CDC: Centers for Disease Control and Prevention.

p < .05; **p < .01; ***p < .001.

8. Discussion

Scarcity of information is often not an issue during a global disease outbreak like the COVID-19 pandemic. On the contrary, individuals are usually bombarded with information that differs exceedingly in its credibility, usefulness, relevance, and authenticity. To better cope with the pandemic, laypeople need to process information in ways that help guide informed judgments and decisions. Such information processing likely involves the reliance on various evidence types. Extending the RISP model in multiple ways via a two-wave survey, we examined the predictors and consequences of reliance on four evidence types during the processing of COVID-19 information. Overall, some RISP variables such as information insufficiency and perceived information gathering capacity consistently predicted the use of all four evidence types. However, other RISP variables, for instance, informational subjective norms, did not emerge as important predictors. In addition, different evidence types had different levels of associations with disease prevention behaviors and satisfaction with the US government’s action. Finally, discrete emotions varied in their influences on the use of evidence types, behaviors and satisfaction.

The first major contribution of this study is the differentiation between different evidence types used during information processing. Unlike other HSM-based research, we focused on what information is utilized rather than how much effort is spent during information processing. Our intent is not to make claims about which type of evidence is superior. Rather, our goal is to test the generalizability of the RISP model in predicting the use of evidence types. Previous research has largely supported the efficacy of the RISP model in explaining systematic processing, but not heuristic processing (Yang et al., 2014). Our findings showed mixed support for the RISP model. Since the four evidence types can be all considered as mental shortcuts or heuristic cues, the focus on them is akin to the focus on heuristic processing, which the RISP model is less suitable to explain. The positive relationships between information insufficiency and all four evidence types suggest that individuals were more likely to use all kinds of evidence when they were motivated by accuracy goals, regardless of the amount of effort required to process certain evidence types. Similarly, the positive relationships between perceived information gathering capacity and four evidence types show that perceived ability to obtain and process relevant information enabled individuals to use all types of evidence in judgment and decision-making. Surprisingly, while a meta-analysis of RISP studies has shown a key role played by informational subjective norms (Yang et al., 2014), this variable was not associated with any evidence type. This finding may simply suggest the minimum impacts of informational subjective norm on the evidence types, especially when the reliance on the four evidence types may be more similar to heuristic processing.

There are also variables associated with the use of an evidence type. For instance, trust in both Chinese and US scientists was associated with reliance on expert evidence. This result suggests that scientists were considered as experts in the domain of COVID-19 regardless of their country of origin. However, overall, we observed a limited role played by trust in relevant channels in influencing the use of evidence types, which might be because we examined only the main effect of trust in relevant channels rather than its interaction effects with information insufficiency and perceived information gathering capacity, which are how relevant channel beliefs are supposed to influence information processing based on the RISP model (Griffin et al., 1999).

The second major contribution of this study is the linkage between the RISP model and subsequent attitudes and behaviors. Despite the connections between information processing and TPB-related variables proposed in the original RISP model (Griffin et al., 1999), the “so what” question remains under-explored in empirical RISP research (Griffin et al., 2013; Lu et al., 2020). Our two-wave study showed that increased use of experiential and expert evidence was related to increased disease prevention behaviors, whereas increased reliance on scientific evidence was associated with decreased satisfaction. Interestingly, the use of the two evidence types that were considered as exemplifying heuristic processing (Trumbo, 1999) had more potential to improve health. This may suggest that personal experience with the disease and experts’ recommendations on what actions to take were more influential in the adoption of health behaviors than statistics or scientific information. As for satisfaction, perhaps when the attention was paid to scientific facts that were arguably more objective than other evidence types, individuals were more likely to realize that the US government has not done enough to address the pandemic. Admittedly, more research is needed to examine the validity of these explanations, which are speculative at best.

We considered the intercultural dimension of the COVID-19 pandemic. While previous research has applied the RISP model to other intercultural risk contexts (e.g. Lu, 2015), our study was the first one to incorporate variables that reflected this aspect. Because we focused specifically on US-dwelling Chinese, we included relevant beliefs regarding both US and Chinese channels, and their satisfaction with the US government’s action to address the pandemic. Intriguingly, we found that trust in US and Chinese channels played somewhat different roles in influencing behaviors and satisfaction. While trust in Chinese channels overall did not influence the adoption of health behaviors, more trust in the US CDC and scientists was associated with reduced health behaviors. These findings might be partly explained by the differences in how the two countries reacted to the pandemic. For example, China has recommended protective behaviors, such as wearing face masks, since the start of the pandemic, whereas it was until much later that the United States made the same recommendation (Lynteris, 2020). Such differences could lead to differing behaviors depending on which country’s channels were trusted more by individuals. Furthermore, research has also documented how US and Chinese media’s coverage of each other is often driven by hostile nationalized ideologies (Bie and Billings, 2015), which could offer an additional explanation for decreased satisfaction in the US government when there was more trust in the Chinese media. These findings illustrate the importance of conceptualizing relevant channel beliefs from an international or intercultural angle whenever possible rather than just a modality (e.g. mediated vs interpersonal, traditional media vs social media) angle.

The third major contribution of this study is the investigation of the influence of discrete emotions on information processing. Adopting the ATF, we incorporated seven discrete emotions into the RISP model. We found that disgust, compassion, anger, and hope played a part in influencing the reliance on evidence types. Due to the exploratory nature of these analyses, it is unclear why these four emotions influenced the use of an evidence type. More investigation is needed to unravel the underlying mechanism. Overall, the ATF appears a suitable framework that could fill the theoretical gap in the original RISP model regarding the role played by affective responses in RISP (Griffin et al., 1999). In addition to its assumptions about how discrete emotions may influence depth of processing based on appraisal dimensions, the ATF also proposes that discrete emotions can influence decisions via the content of thought (Han et al., 2007). For instance, feeling fearful versus angry regarding a risk may put the perceiver into very different mindsets and create desires for different information (Nabi, 2003). It is possible to generate predictive hypotheses regarding these relationships based on the ATF once the emotions and the outcomes (e.g. the depth of processing, the quantity of information seeking, the content of information seeking) are determined. Based on these theoretical groundings, we call for amendments made to the RISP model and direct linkages added between affective responses and information seeking and processing. It is also important to take into account the uniqueness of the RISP model when reconsidering the role played by emotions because the direct relationships between emotions and information seeking and processing have been examined in non-RISP studies (Ferrer et al., 2016). Since information insufficiency is a key component of the RISP model, one promising future direction is to examine how different emotions may influence not only the amount of information that is needed but also what information is needed.

When it comes to attitudes and behaviors, the adoption of disease prevention behaviors was positively related to fear and compassion, but negatively associated with sadness. The fear appeal literature could explain the effects of fear in this case (Tannenbaum et al., 2015). The opposite role played by compassion and sadness is worth noting. Both compassion and sadness can be experienced in response to another’s suffering, which makes it reasonable to assume that both emotions should increase health protection behaviors because for contagious diseases like the COVID-19, protecting oneself also has the benefit of protecting others. However, the negative influence of sadness on behaviors may reflect a different function of this emotion which is associated with retreating from the outside world and staying in an inactive state (Frijda, 1986). Therefore, the sadder individuals felt, the less likely they were going to take action to protect themselves. In terms of satisfaction, we found its negative relationship with anger and positive relationships with elevation and hope. The feeling of anger may come from the perceived incompetence of the US government in addressing the pandemic or from perceived discrimination against overseas Chinese due to coronavirus reportedly originating from China, both of which can fuel dissatisfaction with the US government. In contrast, elevation and hope represent a more positive outlook on the current and future situation, which are unsurprisingly related to more satisfaction.

As with any research, our study is not without limitations. First, despite being relatively large, our sample, which was skewed toward younger, better educated, and mostly sojourning groups, was not representative of all US-dwelling Chinese (U.S. Census Bureau, 2018). Future research may seek to employ a more representative sample, though it would be difficult to obtain a representative sample for a special population like ours. Second, while we adopted a two-wave design which was rarely used for RISP research, the RISP predictor measurements were cross-sectional at Wave 1 and the three outcome variables were cross-sectional at Wave 2. Thus, the directionality of some of the relationships should still be interpreted with caution. Relatedly, based on the HSM, the RISP model proposes that systematic processing is expected to result in more stable attitudes and behaviors (Griffin et al., 1999). However, our measures of behaviors and satisfaction were included only at Wave 2, and we did not have more direct measures of systematic and heuristic processing, which have limited our ability to make claims about the influence of information processing styles on the stability of attitudes and behaviors across time. Future research should conduct a direct test of this proposition, which has not been examined in past RISP studies. Third, due to the large number of variables included in this survey, we used single-item measurements for many variables, especially for discrete emotions, which might reduce the reliability of these measurements. However, research has found little differences between single-item versus multiple-item measurement scales (Gardner et al., 1998), and the use of single items is not uncommon when measuring discrete emotions (e.g. Smith and Leiserowitz, 2014). Finally, there appeared to be some mismatch among our measures. Specifically, some of the measures might relate more to personal risk decisions, whereas others concerned judgments about the virus in general. For instance, the experiential and expert types are more representative of the former, while the statistical and scientific types align more with the latter, which might explain why experiential and expert types were associated with disease prevention behaviors which were more personal but not satisfaction which was more impersonal and why reliance on the statistical type was related to satisfaction but not disease prevention behaviors. In addition, another reason we found a limited role played by informational subjective norms might be that the question item was worded in a general way that could be interpreted as including several dimensions of the disease (e.g. its incidence, governmental responses), which could make it more difficult for the participants to report their response precisely. Future research should attempt to ensure the consistency of their measures and proactively consider the potential impacts when mismatches among measures are unavoidable.

Finally, the findings from this study can provide potentially valuable practical implications for media practitioners, educators, and public health professionals. If the goal is to increase reliance on any evidence type, efforts should be made to increase information insufficiency and perceived information gathering capacity. If the goal is to increase the use of an evidence type, the corresponding predictors should be stimulated. This is also true for motivating the adoption of disease prevention behaviors and changing satisfaction levels. As for the international or intercultural aspect, guidelines on helping immigrant populations navigate through information channels from multiple countries should be developed. These practical goals can be achieved by enhancing science literacy education in schools and to the general public, emphasizing a particular evidence type in media coverage and public health guidelines, and improving media literacy programs in agencies and organizations serving immigrants.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Hang Lu

Haoran Chu

Yanni Ma

Notes

Author biographies

Hang Lu, PhD (Cornell University), is an Assistant Professor of Media Psychology in the Department of Communication and Media at the University of Michigan. His primary research focuses on examining the dynamic role that emotions play in influencing environmental, health, science, and risk communication. In particular, he studies emotions as persuasive appeals, dynamic processes, and situational cues, with a recent focus on the experience of multiple emotions.

Haoran Chu, PhD (University at Buffalo), is an Assistant Professor of Public Relations at Texas Tech University. His research examines the influences of psychological distance, mental construal, and social interaction on people’s response to health and environmental risks. He also studies means to foster individual and community resilience against natural or man-made disasters.

Yanni Ma, PhD (Washington State University), is an Assistant Professor of Organizational and Environmental Communication at Oregon State University. Her research interests include science, environmental, and risk communication.

References

Ajzen

(1991) The theory of planned behavior. Organizational Behavior and Human Decision Processes 50: 179–211.

Asian Pacific Policy and Planning Council (2020) In One Month, STOP AAPI HATE Receives Almost 1500 Incident Reports of Verbal Harassment, Shunning and Physical Assaults. Los Angeles, CA: Asian Pacific Policy and Planning Council.

Bie

Billings

(2015) “Too good to be true?”: US and Chinese media coverage of Chinese swimmer Ye Shiwen in the 2012 Olympic Games. International Review for the Sociology of Sport 50: 785–803.

Chaiken

Liberman

Eagly

(1989) Heuristic and systematic information processing within and beyond the persuasion context. In: Uleman

Bargh

(eds) Unintended Thought. New York, NY: Guilford Press, pp. 212–252.

Eagly

Chaiken

(1993) The Psychology and Attitudes. Fort Worth, TX: Harcourt Brace & Jovanovich.

Ferrer

Klein

Lerner

Reyna

Keltner

(2016) Emotions and health decision-making: Extending the appraisal tendency framework to improve health and healthcare. In: Roberto

Kawachi

(eds) Behavioral Economics and Public Health. Cambridge, MA: Harvard University Press, pp. 101–131.

Frijda

(1986) The Emotions. Cambridge: Cambridge University Press.

Gardner

Cummings

Dunham

Pierce

(1998) Single-item versus multiple-item measurement scales: An empirical comparison. Educational and Psychological Measurement 58: 898–915.

Griffin

Dunwoody

Neuwirth

(1999) Proposed model of the relationship of risk information seeking and processing to the development of preventive behaviors. Environmental Research 80: S230–S245.

10.

Griffin

Dunwoody

Yang

(2013) Linking risk messages to information seeking and processing. In: Salmon

(ed.) Communication Yearbook 36. New York, NY: Routledge, pp. 323–362.

11.

Griffin

Yang

ter Huurne

Boerner

Ortiz

Dunwoody

(2008) After the flood: Anger, Attribution, and the seeking of information. Science Communication 29: 285–315.

12.

Haidt

(2003) The moral emotions. In: Davidson

Scherer

Goldsmith

(eds) Handbook of Affective Sciences. Oxford: Oxford University Press, pp. 852–870.

13.

Han

Lerner

Keltner

(2007) Feelings and consumer decision making: The appraisal-tendency framework. Journal of Consumer Psychology 17: 158–168.

14.

Harring

(2013) Understanding the effects of corruption and political trust on willingness to make economic sacrifices for environmental protection in a cross-national perspective. Social Science Quarterly 94(3): 660–671.

15.

Hogg

(2000) Subjective uncertainty reduction through self-categorization: A motivational theory of social identity processes. European Review of Social Psychology 11: 223–255.

16.

Hornikx

(2005) A review of experimental research on the relative persuasiveness of anecdotal, statistical, causal, and expert evidence. Studies in Communication Sciences 5(1): 205–216.

17.

Hovick

Freimuth

Johnson-Turbes

Chervin

(2011) Multiple health risk perception and information processing among African Americans and whites living in poverty. Risk Analysis 31(11): 1789–1799.

18.

Jeong

Songer

(2008) Understanding scientific evidence and the data collection process: Explorations of why, who, when, what, and how. In: Petroselli

(ed.) Science Education Issues and Developments. New York, NY: Nova Science Publishers, pp. 179–199.

19.

Kahlor

Dunwoody

Griffin

Neuwirth

(2006) Seeking and processing information about impersonal risk. Science Communication 28: 163–194.

20.

Kelley

(1999) Subjective experience as basis of ‘‘objective” judgments: Effects of past experience on judgments of difficulty. In: Gopher

Koriat

(eds) Attention and Performance XVII: Cognitive Regulation of Performance. Interaction of Theory and Application. Cambridge, MA: MIT Press, pp. 515–536.

21.

Kline

(2016) Principles and Practice of Structural Equation Modeling, 4th edn. New York, NY: Guilford Press.

22.

Lazarus

(1991) Emotion and Adaptation. New York, NY: Oxford University Press.

23.

(2015) Burgers or tofu? Eating between two worlds: Risk information seeking and processing during dietary acculturation. Health Communication 30: 758–771.

24.

, APPC 2018–2019 ASK Group, Winneg

Jamieson

Albarracín

(2020) Intentions to seek information about the influenza vaccine: The role of informational subjective norms, anticipated and experienced affect, and information insufficiency among vaccinated and unvaccinated people. Risk Analysis 40(10): 2040–2056.

25.

Lynteris

(2020) Why do people really wear face masks during an epidemic? To fend off disease, but also to show solidarity. The New York Times. Available at: https://www.nytimes.com/2020/02/13/opinion/coronavirus-face-mask-effective.html

26.

Nabi

(2003) Exploring the framing effects of emotion: Do discrete emotions differentially influence information accessibility, information seeking, and policy preference? Communication Research 30: 224–247.

27.

Rieke

Sillars

(1984) Argumentation and the Decision Making Process. Hoboken, NJ: Wiley.

28.

Schemer

Matthes

Wirth

(2008) Toward improving the validity and reliability of media information processing measures in surveys. Communication Methods and Measures 2: 193–225.

29.

Smith

Leiserowitz

(2014) The role of emotion in global warming policy support and opposition. Risk Anal 34(5): 937–948.

30.

Tannenbaum

Hepler

Zimmerman

Saul

Jacobs

(2015) Appealing to fear: A meta-analysis of fear appeal effectiveness and theories. Psychological Bulletin 141: 1178–1204.

31.

Tonelli

(2006) Integrating evidence into clinical practice: An alternative to evidence-based approaches. Journal of Evaluation in Clinical Practice 12(3): 248–256.

32.

Trumbo

(1999) Heuristic-systematic information processing and risk judgment. Risk Analysis 19: 391–400.

33.

Trumbo

McComas

(2003) The function of credibility in information processing for risk perception. Risk Analysis 23: 343–353.

34.

U.S. Census Bureau (2018) Selected Population Profile in the United States, 2018 American Community Survey 1-Year Estimates. Washington, DC: U.S. Census Bureau.

35.

Yang

Aloe

Feeley

(2014) Risk information seeking and processing model: A meta-analysis. Journal of Communication 64: 20–41.

36.

Yang

Chu

Kahlor

(2019) Fearful conservatives, angry liberals: Information processing related to the 2016 presidential election and climate change. Journalism & Mass Communication Quarterly 96: 742–766.

37.

Yang

McComas

Gay

Leonard

Dannenberg

Dillon

(2010) From information processing to behavioral intentions: Exploring cancer patients’ motivations for clinical trial enrollment. Patient Education and Counseling 79: 231–238.

38.

Yuan

(2020) China in coronavirus propaganda push as US ties worsen. Al Jazeera English. Available at: https://www.aljazeera.com/news/2020/3/26/china-in-coronavirus-propaganda-push-as-us-ties-worsen