1 INTRODUCTION

One of the hallmarks of the COVID-19 pandemic has been the politicization of facts, with high levels of disagreement on the risks associated with COVID-19 between political parties and ideologies. This rejection of science along party lines is nothing new, as can be seen by the extensive research on how political ideology is associated with acceptance of climate change (Kahan et al., 2012). In part, the seeming dismissal of science during the COVID-19 pandemic is due to the conflicting information about the risks associated with the virus from different sources and widespread use of social media to further spread misinformation (Huynh, 2020a). For example, when cases started to rise and warnings were being issued by the Center for Disease Control and Prevention (CDC), President Trump downplayed the threat of the virus, saying it was under control and not as dangerous as the flu (Woodward, 2020).

In the wake of such extreme differences in risk assessment and spreading of misinformation between people, many researchers sought to characterize factors that can influence our understanding of the virus and its associated risk (Van Bavel, Baicker, et al., 2020). The importance of understanding these factors was further underlined by data collected from 10 countries, which found that COVID-19 risk perception correlated significantly with endorsement of preventative health behaviors (Dryhurst et al., 2020). Most studies aimed at understanding these individual differences factors have focused on political ideology, personality traits, and demographic information such as nationality or gender (Abdelrahman, 2020; Gerhold, 2020; Huynh, 2020b). Of note, a recent theoretical overview by Van Bavel, Baicker, et al. (2020) highlights much of the research on how social norms, inequality, political polarization, and other factors have influenced our response to the pandemic. However, there is a largely untapped body of literature on cognitive abilities that can influence risk assessment (e.g., Reyna, 2020; Reyna et al., 2009). Some researchers have found that cognitive sophistication, knowledge of the virus, current events, and medical training influenced preventative behaviors and a better understanding of the virus, and risk assessment (Motta Zanin et al., 2020; Pennycook et al., 2020; Pennycook & Rand, 2020; Taghrir et al., 2020). However, others have disputed the importance of knowledge and cognitive ability on assessing risk and misinformation, showing that variables such as politics and religiosity play a larger role (Allum et al., 2008; Kahan et al., 2012; Lewandowsky & Oberauer, 2016).

Given the importance of understanding factors that can influence how people assess risk and misinformation, the lack of research on cognitive variables, and the mixed findings in literature, we endeavored to clarify the importance of key cognitive variables in how people understood the virus and its risks. Following Fuzzy-Trace Theory (Reyna, 2008) our conceptual model is that COVID-19 related behavioral intentions are informed by gist representations of risk (Reyna, 2012, 2021) and risk thresholds for specific actions. Individual differences in cognition and political affiliation are hypothesized to affect those representations. We selected subjective numeracy, cognitive reflectiveness, and impulsivity as our key cognitive variables. Each has been shown to influence our ability to assess medical risk or reject misinformation (Pennycook et al., 2015, 2020; Peters et al., 2006; Zimmermann, 2010). However, given the politicization of facts and the widespread misinformation coming from sources of authority, it is unclear how these variables would relate to COVID-19 specifically in the context of political ideology.

Numeracy (i.e., literacy for quantitative concepts) has been identified as contributing to individual differences health risk perception. Subjective numeracy (Fagerlin et al., 2007), which refers to an individual’s perceptions of their own quantitative skills, appears to be an appropriate measure for risk perceptions about COVID-19 given its importance in medical judgment and decision making (Liberali et al., 2012; Peters et al., 2006). For example, participants with higher numeracy tend to be less influenced by presenting information in misleading ways and are generally better able to understand risk outcomes (Peters et al., 2006; Reyna et al., 2009). Numeracy clearly plays a role in health risk perception, yet numbers are surprisingly ambiguous, yielding the potential for large differences in interpreting the same risk-related data (Reyna, 2020).

Cognitive reflectiveness, as measured by the Cognitive Reflection Test (CRT; Toplak et al., 2014) refers to an individual’s ability to avoid incorrect intuitive responses by being more thoughtful and reflective in their judgment and reasoning process. While related to cognitive ability and rational thinking styles, cognitive reflection scores seem to primarily indicate a level of cognitive miserliness, with higher scores reflecting a greater dedication of cognitive resources toward a task (Toplak et al., 2014). Higher CRT scores predict judgment and reasoning ability across a wide variety of tasks (Frederick, 2005; Hoppe & Kusterer, 2011), and lower endorsement of COVID-19 misperceptions (Pennycook et al., 2020), perhaps due to enhanced deliberation of the false statements (Van Bavel, Baicker, et al., 2020). Given the mathematical nature of the CRT, some have suggested that much of the power of the CRT in predicting decision making relies on numerical ability (Sinayev & Peters, 2015). However, other research indicates that cognitive reflectiveness contributes its own unique benefit to judgment and decision making tasks, above and beyond numeracy (Liberali et al., 2012). These findings suggest that in addition to assessing numeracy, a measure cognitive reflectiveness may be beneficial for both assessing medical information and medical risk, though research on the importance of cognitive reflectiveness in medical judgment and reasoning is still in its infancy.

Another related individual difference under investigation is impulsivity. It is generally agreed upon that impulsivity is a construct that encompasses many executive functions including delayed discounting, reward evaluation, suppression of responses, and focusing of attention. More broad definitions of impulsivity tend to include behaviors that are unnecessarily risky, premature, or are poorly conceived (De Wit, 2009). Greater impulsivity has been specifically linked to higher risk-taking behavior as it concerns personal health, such as in the use of drugs and alcohol, unhealthy eating habits, or high-risk sexual activity (Braddock et al., 2011; Dawe & Loxton, 2004), and has recently been linked to lower perceptions of health-related risks for substance abuse (Zimmermann, 2010). As such, it presents a promising factor in an individual’s ability to assess medical risk and misinformation during the COVID-19 pandemic.

Political ideology is hypothesized to influence risk assessment with respect to COVID-19 in three ways. First, beyond the scope of our investigation, ideology appears to determine sources of information about COVID-19 in the form of news, web sites, social media, etc. Second, political affection is hypothesized to affect participants’ perceptions of the risks associated with the pandemic (Barrios & Hochberg, 2020; Calvillo et al., 2020). Finally, ideology is hypothesized to influence thresholds for acceptable risk. Thus, we predict that liberals will be more likely than conservatives to consider actions such as being in a crowded room risky and also have a lower threshold for acceptable risk.

2 STUDY 1

We conducted two studies during the peak of the pandemic. To date, individual responses to the COVID-19 pandemic have mostly been investigated in the context of social constructs and personality traits. However, little is known about the importance of cognitive reflectiveness, subjective numeracy, and impulsivity in assessing COVID-19 risk or misinformation. In this study, we sought to expand upon previous findings on the importance of political ideology and cognitive reflectiveness in assessing misinformation and risks for COVID-19, and investigated the role subjective numeracy, impulsivity, and the presence of a risk factor for COVID-19 play in risk estimation, risk aversion, and their willingness to endorse common misperceptions related to COVID-19.

3 METHOD

4 RESULTS AND DISCUSSION

5 STUDY 2

Given the promising results of Study 1, the focus of Study 2 was to find the best predictors of healthy behavioral intentions, such as wearing a mask and getting vaccinated, and to test the relative persuasiveness of different arguments about COVID-19. Argumentation has been found to be effective in promoting healthy decisions and behaviors under some circumstances (e.g., Cedillos-Whynott et al., 2016) and are also sensitive to individual differences (Wolfe, 2012). An argument is, at minimum a claim supported by one or more reasons (Wolfe et al., 2009). Previous research suggests that when people agree with an argument, they are mostly agreeing with the claim, whereas when someone says that an argument is strong or weak, they are primarily weighing the supporting reasons (Wolfe et al., 2009). This suggests, contrary to notions of rational persuasion, that persuasion through written arguments is challenging because it is possible to admit that reasons supporting opposing beliefs are strong without changing one’s level of agreement. This is especially challenging in partisan and ideological contexts where claims and reasons may be processed through differently weighted argumentation schema (Wolfe et al., 2009) and subject to different biases (Wolfe & Britt, 2008). Thus, argumentation has the potential to be a valuable tool in promoting health behaviors related to vaccination, but there are numerous challenges identified in the literature.

6 METHOD

7 RESULTS AND DISCUSSION

7.1 Perceived risk

As seen in Study 1, high risk participants had higher risk estimates for being hospitalized with COVID-19, t(87.00) = 3.82, p < .001, d = 0.63, dying from COVID-19, t(86.65) = 2.76, p = .001, d = 0.45, and developing serious complications, t(88.36) = 3.00, p = .003, d = 0.48, as can be seen in Table 3. High risk participants were not significantly different in their risk estimates for being infected with COVID-19, t(115.94) = −0.52, p = .603, d = −0.07, spreading the disease to friends or family members, t(349) = −0.31, p = .759, d = −0.04, spreading to other students, professors, or staff members, t(345) = −1.19, p = .236, d = −0.16, or killing someone by spreading the virus, t(342) = .69, p = .491, d = 0.09. Unlike Study 1, having a risk factor did not impact participants’ estimates for being infected with COVID-19 while engaged in specific activities such as going shopping or getting a haircut, t(354) = −1.09, p = .272, d = −0.15, perhaps due to this study taking place later in the pandemic when more people had a better understanding of the risks of infection.

TABLE 3.
Mean (standard deviation) perceived risks associated with Covid-19 by presence of risk factor(s) in Study 2

Perceived COVID-19 risk	Participants with risk factors	Participants without risk factors
General risk
Infection	36.09 (23.71)	37.78 (26.57)
Hospitalization	27.4 (27.88)	13.88 (19.48)***
Dying	12.77 (16.29)	6.94 (12.10)***
Serious long-term complications	27.01 (25.68)	17.09 (19.20)**
Spreading to friends or family members	46.99 (29.45)	48.21 (29.68)
Spreading to students, faculty, or staff	40.07 (29.16)	44.93 (30.96)
Killing someone else through spreading	27.84 (27.89)	25.33 (26.89)
Activity-related risk of infection
Grocery shopping	25.67 (22.11)	26.26 (22.25)
Getting a haircut	22.93 (22.00)	24.57 (21.70)
Hanging out with fewer than 10 people	24.34 (20.11)	27.64 (20.75)
Hanging out with more than 10 people	44.44 (26.09)	49.57 (24.35)
Going to the gym	34.43 (23.25)	36.20 (24.77)
Travel in the United States	40.61 (26.85)	45.10 (25.38)
International travel	47.53 (28.17)	49.97 (26.41)
Going to a restaurant	28.29 (21.66)	31.55 (22.64)
Going to the beach	26.80 (22.98)	29.33 (24.41)

• Note: *p < .05, **p < .01, ***p < .001.

Looking at the same cognitive individual differences measures as in Study 1, cognitive reflectiveness and impulsivity were independent of each other once again, r(365) = −.03, p = .584. Subjective numeracy was negative correlated with impulsivity, r(365) = −.19, p < .001, and positive correlated with CRT scores, r(365) = .33, p < .001. We simultaneously regressed our three individual differences measures onto general perceived risk, and found that our model was significant, F(4,362) = 3.21, p = .023, R² = .03. As before, impulsivity, β = 0.08, t = 1.54, p = .125 and subjective numeracy, β = 1.31, t = 0.82, p = .414, did not significantly predict participants estimates of general perceived risk surrounding COVID-19. However, CRT scores were a significant negative predictor of general perceived risk estimates, β = −0.15, t = −2.72, p = .007, suggesting that more reflective participants tended to have lower risk estimates. Looking at activity related risk, our model was also significant, F(4,364) = 3.98, p = .008, R² = .03. CRT scores, β = −0.15, t = 2.75, p = .006 but not impulsivity, β = .09, t = 1.73, p = .084, or subjective numeracy, β = .001, t = 0.03, p = .980, predicted lower activity-related risk perception. Together, these results suggest that cognitive reflectiveness may reflect a more appropriate perception of risk as the populace’s understanding of the virus improved over time.

Using a one-way ANOVA, we also assessed the role political ideology had on participants’ perceived risks, and found that activity-related perceived risk, F(2, 360) = 8.79, p < .001, η² = .05, but not general perceived risk, F(2, 360) = 1.76, p = .173, η² = .01, differed by political ideology (see Figure 2). Bonferroni corrected pairwise comparisons revealed that conservative participants rated risks as significantly lower than liberal (p < .001) and moderate (p = .001) participants, replicating our early finding.

7.5 Behavioral intentions

Simultaneously regressing the three behavioral intentions onto our three individual differences revealed that our models for vaccination, F(3, 362) = 3.51, p = .015, R² = .03, and social distancing, F(3, 364) = 3.75, p = .011, R² = .03, were significant, but our model for wearing a mask was not, F(3, 364) = 2.45, p = .064, R² = .02. Cognitive reflectiveness significantly predicted greater likelihood for getting vaccinated, β = 0.15, t = −2.73, p = .007, wearing a mask, β = 0.14, t = 2.52, p = .012, and maintaining social distancing, β = 0.18, t = 3.28, p = .001. Impulsivity was not a significant predictor for getting vaccinated, β = −0.06, t = −1.19, p = .234, wearing a mask, β = 0.002, t = 0.04, p = .965, and maintaining social distancing, β = −0.04, t = 0.82, p = .415. Similarly, subjective numeracy did not predict getting vaccinated, β = −.01, t = 0.18, p = .858, wearing a mask, β = −0.10, t = 1.73, p = .085, and maintaining social distancing, β = −0.06, t = −1.03, p = .304. As can be seen in Figure 3, political ideology played a significant role in participants’ intention to get vaccinated, F(2, 360) = 32.72, p < .001, η² = .16, to wear a mask, F(2, 362) = 53.93, p < .001, η² = .23, and to maintain social distancing, F(2, 362) = 54.18, p < .001, η² = .23. Bonferroni-corrected pairwise comparisons revealed that in all cases, liberal participants had the highest ratings, followed by moderate participants, and conservative participants, respectively. All pairwise comparisons were significant with an alpha of .05.

8 GENERAL DISCUSSION

In this work, we aimed to better understand what leads people to engage in risky behaviors, inaccurately assess risk, and believe in common misperceptions surrounding COVID-19. We found that across both studies participants with at least one risk factor had higher risk estimates for hospitalization, complications, and death from COVID-19. Conservative participants rated COVID-19 risks as lower, were more willing to accept activity-related risk, and were more likely to believe misperceptions. In Study 2, conservative participants were less likely to wear a mask, socially distance, or get vaccinated. Impulsive participants tended to believe more misperceptions, whereas reflective individuals were more risk-averse and more likely to engage in preventative health behaviors. We also found that regardless of political ideology, and consistent with previous research (Wolfe et al., 2009), when agreeing with arguments, participants place more weight on an argument’s claim than the reason behind that claim. However, our findings with respect to argument strength and supporting reasons differed from prior research. Overall, these findings present how multiple individual differences, including cognitive, political, and personal factors, can impact a person’s response to the COVID-19 pandemic.

Across both studies, we found that having a COVID-19 risk factor is consistently associated with increased risk assessment of complications, hospitalization, and death, supporting the importance of personalizing and contextualizing facts and statistics on the virus. Indeed, research suggests that people give considerable weight to their own experiences over and above aggregate statistical information in assessing and making decisions about medical risk (Holmberg et al., 2015). The events themselves were secondary to the way people experienced those events and were given more weight than statistical information. Thus, the stories people tell themselves about their own risk factors are likely to strongly contribute to their own gist representations (Reyna, 2008) and play a significant role in risk perception.

Given the extreme politicization of COVID-19 in the United States, political attitudes seem to have a strong influence on perceptions of risk toward the pandemic. In general, political ideology can influence several cognitive processes related to risk perception including belief bias in causal reasoning (Gugerty & Shreeves, 2020). However, increasing political polarization in the United States has led to contrasting views and interpretations of events beyond what we might normally expect. In this study, we replicated previous findings on the influence political ideology can have on participants’ perception of risk for the pandemic, showing a clear negative relationship between risk perception and conservatism (Barrios & Hochberg, 2020; Calvillo et al., 2020). However, rather than assessing general worry or emotional responses, as most studies have done, we found that conservatives’ actual risk estimations were lower while their estimates for acceptable maximum risk were higher. Taken in conjunction with other research, our results provide a more holistic understanding of how politics can influence risk perception and the acceptability of said risk.

Similarly, compared to liberal participants, both moderate and conservative participants believed more misinformation about the pandemic. Research has shown that people become susceptible to misinformation when the world does not make sense to them (Reyna, 2020). The pandemic has been the subject of several conspiracy theories directed at conservatives, often from popular news networks or the former president himself. Importantly, it is not necessary for people to “believe” conspiracy theories for them to influence risk assessment. Research on argumentation suggests that people are more likely to agree with claims supported by blatantly bogus reasons and warrants than unsupported claims (Wolfe et al., 2018). This effect holds true even when contradicting domain specific knowledge for relatively knowledgeable individuals (Wolfe et al., 2018). Because conspiracy theories provide reasons, they can serve to satisfy the cognitive constraints imposed by schema, they may influence risk perception without being judged true or credible. Research suggests that if a claim is neutral (or one a person is likely to support), even utterly ridiculous reasons increase agreement; most any reason is better than none (Wolfe et al., 2018).

One of the key findings of the present work was the role cognitive reflection played in assessing risk and misinformation. Our work replicated previous findings on the importance of cognitive reflection on not being swayed by misinformation (Pennycook et al., 2020; Pennycook & Rand, 2019). We also found that higher cognitive reflection predicted lower levels of maximum acceptable risk. The standard interpretation of these results is that participants who are more reflective and analytical are better able to navigate the current deluge of misinformation surrounding the pandemic and are more risk averse. However, performance on the CRT may have alternative explanations. For example, several studies, including ours, have found a significant relationship between cognitive reflectiveness and numeracy scores (Patel et al., 2019; Sinayev & Peters, 2015). This link between cognitive reflectiveness and numeracy is further supported by the fact that the items in the CRT all require some mathematical ability and knowledge. However, other work suggests that measures of numeracy and cognitive reflection are tapping overlapping but inherently different processes important for judgment and reasoning, (Liberali et al., 2012), a finding which can be partially supported by participants’ similar performance on non-mathematical CRT versions (Campitelli & Labollita, 2010). Our findings support the latter conclusion, as numeracy and cognitive reflectiveness predicted different responses to COVID-19 risk and misinformation assessment. However, future work should continue to explore the specific roles reflective processes and numerical ability play in participants’ responses to the pandemic.

In line with our findings of cognitive reflection, lower levels of impulsivity predicted fewer endorsements of misperception, but did not relate to risk assessment or risk acceptability. How impulsivity relates to risk-perception or acceptance seems to be domain-specific. For example, the relationship between impulsivity and risk perception is consistent in the context of drugs and alcohol, but less so in the context of high-risk sexual activity where risk perception can be more complex (Mehrotra et al., 2009). It is possible that the politicization and overabundance of misinformation on COVID-19 risks muddies the waters, making this a context where assessing risk is less straightforward, and thus less clearly related to impulsivity.

Numeracy has long been known to be associated with judgment and decision making related to health (Lipkus & Peters, 2009). However, our hypotheses about numeracy were not supported by the data. It may be that the politicized nature of COVID-19 overwhelms numeracy as a predictive individual difference. Citing Kahan et al. (2012), Lewandowsky and Oberauer (2016) report that with materials demonstrating the base rate fallacy that are problem isomorphs, numeracy was a good predictor of performance on a problem related to the effectiveness of a skin cream, but not for a numerically identical problem involving the politicized issue of gun control. In the case of gun control, political ideology predicted performance and numeracy did not. As Reyna (2020 p. 672) notes, “numeracy is not sufficient to understand risk. In fact, numbers are ambiguous in the way that words are ambiguous, perhaps more so.”

Looking at our individual differences measures as a whole, these studies suggest that judgments about the level of risk for behaviors, such as going shopping, are distinct from personal thresholds for what constitutes too much risk. While it appears that political ideology affects both judgments on risk perception and acceptability, other variable such as reflectiveness predict lower thresholds for acceptable risk and, at least in Study 2, lower risk perception. While initially counterintuitive, this may simply show that, despite a lower perception of risk, reflective participants are still more risk-averse. However, the joint influence of thresholds and risk perception should be studied in future research.

With respect to argumentation, typically when people agree with an argument they are primarily agreeing with the claim, whereas when they assess the strength or quality of an argument, they are more heavily weighing the supporting reasons (Wolfe et al., 2009). However, in the present research there were ideological differences for judgments of argument strength. While liberal participants weighed claims and reasons evenly, conservatives weighed the claim more heavily. In other words, conservative participants were more likely to base judgments about the strength or quality of arguments on claims such as “It is good that the Federal Government is making vaccines against COVID-19 available to Americans free of charge” rather than on supporting reasons, be they those that pilot testing suggests are relatively conservative, such as “because that was the goal of President Trump’s Operation Warp Speed” or relatively liberal such as “because protecting the health of Americans is a basic function of government.” Unfortunately, these findings highlight the difficulty in developing effective argument-based interventions to promote healthy decisions and behaviors. It appears that in this partisan environment, the argumentation schemata of conservatives and liberals weigh theme and side above warrants and reasons (Wolfe et al., 2009). Instead, there are reasons to believe that a more fruitful approach is to give people the information and conceptual tools they need to understand the qualitative, contextualized meaning of risk information (Reyna, 2020).

Finally, looking at participants’ intent to engage in healthy preventative behaviors, we saw that participants who were liberal and more reflective were more likely to engage in all three behaviors. These findings could be useful in helping address vaccine hesitancy. Despite the fact that adults over the age of 16 have been able to get vaccinated for several months, young adults are still plagued by vaccine hesitancy, citing concerns for the vaccine’s safety and efficacy as their main reason to avoid it (Baak et al., 2021). Many of these concerns can be linked to widespread misinformation, largely associated with conservative circles. This may explain why reflective participants, who can better ascertain what information is valuable, and liberal participants may be more willing to get vaccinated. We may able to help more of these young adults overcome their vaccine hesitancy by directly targeting our efforts on conservatives who are confronted with more misinformation, and providing more consistent, bipartisan messaging that appeals to a sense of national identity (Van Bavel, Cichocka, et al., 2020).

9 CONCLUSIONS

Here, we have taken preliminary steps towards understanding the cognitive individual differences that may cause people to inaccurately assess risks of COVID-19, engage in risky behaviors, and endorse misperceptions. Our findings support the importance of cognitive ability in avoiding misinformation during the current pandemic, and the effect political polarization can have on an individual’s perception of medical risk, willingness to accept risk, tendency to believe misinformation, and intention to engage in healthy behaviors. By clarifying these individual differences, we can further understand health behaviors, and more effectively create interventions targeted at populations that need them most. One limitation of the current work is that represents a “snapshot” in time. However, some social and political influences on risk perception may be dynamic. Additionally, we asked about behavioral intentions and hypothetical actions, but did not assess potentially risky behaviors directly. Thus, it is not clear whether estimates for statements about the risk associated with getting a haircut correspond with the frequency of actually visiting a hair salon. A third limitation is that only about 4% of our sample self-identified as “strong conservative.” These studies lacked the statistical power to make fine-grained distinctions based on ideology, or separate “social conservatives” from “fiscal conservatives” etc.

One policy implication of these studies is that public health interventions designed to reduce the spread of COVID-19 should take ideology into account during development and testing. Future work should continue to expand upon our current knowledge of COVID-19 risk perception. The role of argumentation in influencing risk perception warrants further research, particularly in the context of political ideology and personal risk factors. Now that sufficient time has passed, there is a wealth of knowledge to be gleaned on how individuals differed in their actual behaviors, acceptance of health-related mandates, and willingness to be educated on the pandemic, to name a few key questions. Furthermore, understanding the processes and abilities that shaped the response to the current pandemic can ideally help us prepare for the next one.

CONFLICT OF INTEREST

The authors declare no conflict of interest.