Science audiences, misinformation, and fake news

Knowledge about scientific facts.

The US Science & Engineering Indicators (SEI) surveys measure factual knowledge about science biannually as the average number of correct answers to a series of true-or-false and multiple-choice items. Although such batteries of closed-ended factual knowledge questions are imperfect representations of what citizens know about science more broadly, these trend data suggest that knowledge of scientific facts and terms has not decreased significantly in recent years. There are differences across respondents, however, with factual knowledge being “strongly related to individuals’ level of formal schooling and the number of science and mathematics courses completed” (8).

Epistemic knowledge about science.

Notably, recalling isolated scientific facts might have a limited impact on citizens’ ability to make meaningful policy choices about risks and benefits surrounding emerging technologies. Likely more important is what has been called “epistemic knowledge”—that is, levels of information or misinformation among nonexpert publics about the scientific process and how this process shapes the findings produced by science (9). This is slightly different from discussions of “epistemic beliefs” more broadly, which can be understood as “beliefs about the nature of knowledge and how one comes to know” (10).

Survey data on Americans’ (mis)understanding of the scientific process does not show meaningful changes over time (with some variations due to coding of open-ended data). In the most recent SEI survey, one in three Americans (36%) misunderstood the concept of probability; half of the population (49%) was unable to provide a correct description of a scientific experiment; and three in four (77%) were unable to describe the idea of a scientific study, indicating an inability among “[m]any members of the public … to differentiate a sound scientific study from a poorly conducted one and to understand the scientific process more broadly” (11).

Public opinion surveys also suggest that significant proportions of the public are concerned about how low levels of epistemic knowledge can negatively influence their own or other Americans’ understanding of science news. Although there is certainly some complexity (and irony) in measuring individuals’ perceptions of misperceptions among their peers, it is nonetheless interesting to note that two in five Americans (44%) agree that it is “a big problem” that “[t]he public doesn’t really know enough about science to understand findings in the news,” with only 16% not considering it a problem. Similarly, 4 in 10 (40%) think that it is “a big problem” that “[t]here are so many findings that it’s hard to distinguish between high and low quality studies,” with only 18% not seeing it as a problem (12).

Inaccurate views of scientific consensus and the willful rejection of scientific consensus.

In a 2014 US survey (15), two-thirds of respondents (67%) thought that scientists did “not have a clear understanding about the health effects of GM crops,” despite broad scientific consensus on the topic (16). Similarly, half of Americans (52%) thought scientists were “divided” in the belief that the universe was created in the Big Bang, and about a third each thought that scientists were divided on anthropogenic climate change (37%) and evolution (29%). Of course, these data do not make clear the cause of these inaccurate views, which, arguably, could stem from people being uninformed, intentionally misinformed, or a bit of both.

Furthermore, split-ballot survey experiments have shown that even when Americans do seem to possess accurate knowledge of scientific consensus (however large or small that number may be for a given issue) there is no guarantee that they will integrate that knowledge into their attitudes or policy preferences (17). In other words, these respondents know what the scientific community has established as fact, but they nonetheless refuse to “know” it. Some have therefore argued that rejection of scientifically accurate accounts of the Big Bang or evolution by nonexpert audiences indicates neither a lack of information about scientific consensus nor the presence of misinformation, but, rather, motivated information processing (18).

Conspiratorial beliefs.

Another potential problem is persistent belief in conspiracy theories, or theories in which explanations for events and phenomena offer “as a main causal factor a small group of persons (the conspirators) acting in secret for their own benefit, against the common good” (19). Conspiratorial beliefs can thus involve not only a willful rejection of scientific consensus but also false attributions of intent to members of the scientific community, as well as the fabrication of relationships between actors. For this reason, conspiratorial beliefs are typically understood as distinct from simple ignorance or misperception about isolated facts. Many of us believe in facts that turn out to be wrong. For instance, 7 in 10 Americans falsely attribute the statement “I can see Russia from my house” to Sarah Palin instead of Saturday Night Live’s Tina Fey (20). Many individuals, however, would adjust their views about this fact when presented with information showing their initial beliefs to be wrong. However, individuals who endorse conspiracy theories often refuse to adjust their belief systems when they are confronted with new and better information contradicting their misunderstandings (21).

Importantly, there is evidence to suggest that additional, traditional education will not be enough to dispel belief in conspiracy theories. For example, conspiratorial beliefs and inaccurate beliefs about scientific issues such as vaccine safety and climate change have also been linked to certain “epistemic beliefs,” or broader convictions about how people can and should come to know what is true. Specifically, people who “[put] more faith in their ability to use intuition to assess factual claims than in their conscious reasoning skills” are particularly likely to support conspiracy theories, whereas people who believe that empirical evidence is needed to validate truth claims exhibit the opposite tendency (10). This is true even when issues have become politicized, leading to the conclusion that belief in conspiracies is “[f]ar from being an aberrant expression of some political extreme or a product of gross misinformation [but rather] a widespread tendency across the entire ideological spectrum” (22).

It might be argued that survey respondents who express belief in conspiracy theories do not actually believe the falsehoods, but rather that they endorse certain conspiratorial views as a means of expressing their political or ideological allegiances, or to engage in a sort of out-group “mudslinging” (e.g., by claiming that Barack Obama is not American). Recent research, however, supports the claim that individuals actually do hold the conspiratorial beliefs they assert and are not simply engaging in “expressive survey responding” (23). Unfortunately, conspiratorial beliefs often persist because the falsehoods which help to sustain them are repeated and “boosted” by politicians, corporate actors, fringe media organizations, and others to mobilize political support from their base (7).

The (in)ability to recognize misinformation.

Implicitly, most approaches to algorithmic curation of facts assume that citizens are misinformed because they are unable to sift through and critically evaluate information in emerging (social) media environments. There is no doubt that low levels of media literacy among citizens are part of the problem (25).

News and media literacy has been broadly defined as “the ability to access, analyze, evaluate and create messages in a variety of forms” (26). Arguably, it is the “evaluation” skill that poses the most relevant challenge for misinformation, as those with limited ability to evaluate “cannot distinguish dated, biased or exploitative sources” (26). A recent assessment of American students’ media literacy demonstrates that the vast majority of them struggle to (i) recognize the possible biases of politically charged tweets and (ii) distinguish between a news story and news-like advertisement (27). Moreover, as the Pew Research Center (28) reports, one in four (23%) American adults admitted to sharing misinformation via social media.

These circumstances have led some to argue that “the ultimate check against the spread of rumor, pernicious falsehood, disinformation, and unverified reports masquerading as fact” is a “generation of astutely educated news consumers” who can also function as competent digital content producers, and who can “identify for themselves fact-and-evidence-based news and information” (29). Interestingly, others have critiqued proposals to increase media literacy by noting that these efforts have the potential to backfire, as “some media literacy skills could be used to justify belief in misinformation [and that] elite discourse about ‘fake news’ may decrease the public’s trust in the media and their ability to recognize actual news stories without facilitating the identification of fake news” (30). Consequently, these researchers suggest pairing news media literacy education with “activities designed to spur discussion about political issues.”

Indeed, recent events such as the 2016 presidential election have brought increasing public attention to the role of social media in structuring and presenting information in such a way that may limit an individual’s ability to assess the quality and usefulness of information, and to distinguish between fact and fiction. In response to mounting criticism of this kind, Facebook and others have proposed a dizzying array of technical remedies intended to bolster users’ ability to identify misinformation and, generally, to make it easier for their users to have more “positive” information encounters. Unfortunately, their hasty attempts to diffuse criticism have often backfired or have the potential to backfire, which they have sometimes admitted (31).

For example, Facebook recently announced that it will prioritize the display of content that has been shared by users’ friends and family members, and that they will see “less public content, including news, video and posts from brands,” in an effort to offer users more “meaningful connections,” and to ensure that Facebook is a force for good (32). However, this latter solution may cause people to see “more content that reinforces their own ideologies” (33), and, in countries where these specific technical changes were piloted, frustrated users have reported that the modifications actually promoted the spread of fake news (34).

As technology companies continue to grapple with changes to their algorithms and interfaces, third-party fact-checking groups such as PolitiFact.com and Factcheck.org have also emerged to boost people’s abilities to debunk misinformation (35), and efforts are ongoing to offer reliable, automated “deception detection” for both text and images (36, 37). Furthermore, Google is working on real-time “fact check snippets” that appear as individuals search for disputed information (38), and computer scientists are devising solutions to automatically detect and combat the influence of “bots” (robots), which have been shown to successfully spread fake news with real-world consequences in both politics and the stock market (39). Finally, there is some evidence that correcting misinformation via an algorithmically driven “related stories” function on social media platforms can reduce misperceptions of science-related information (40).

Technical innovations like the ones above have been rightfully advocated for as possible solutions to the spread of misinformation. However, one unfortunate and defining feature of the posttruth era is that “facts and objective evidence are trumped by existing beliefs and prejudices,” such that “a notable segment of the American public now subscribes to a non-standard epistemology that does not meet conventional criteria of evidentiary support” (41). If it is indeed the case that facts no longer matter as much as they normatively should to some Americans, then technical solutions that make facts more recognizable or more visible will need to be supplemented by strategies that pair individual ability-focused solutions with solutions that address individuals’ (lack of) motivation to seek out, consume, and interpret information in ways that privilege accuracy over other possible goals, such as the protection of their preexisting beliefs.

Motivations to recognize inaccurate information … or not.

Beyond issues of ability, there are also psychological factors that contribute to people becoming misinformed, and these factors can make it difficult for people to identify for themselves what is fact versus fiction. Specifically, individuals are more likely to accept information that appears to follow a logical narrative, that comes from a source they perceive to be “credible,” that is consistent with their preexisting values and beliefs, and that seems to be something other people believe (42). The psychological unease or inconvenience of encountering worldview-challenging information can produce a desire to minimize feelings of “cognitive dissonance,” which can lead to biased perception and information processing that complicates the recognition and rejection of falsehoods (43).

Among other strategies, people may grapple with the complexity of the external world by engaging in selective exposure and motivated reasoning. Selective exposure refers to the act of choosing to read or view belief-consistent information over belief-inconsistent information (when given the choice), and there is evidence that such selectivity occurs among strong partisans who are especially knowledgeable about politics and who gravitate more toward news sources that mirror their preexisting views (44). Importantly, selective exposure is not limited to politics and has also been shown to occur, for example, as individuals seek information about scientific topics (45). When individuals expose themselves primarily to media sources that (knowingly or unknowingly) convey falsehoods, there is some evidence that selective exposure plays a role in keeping people misinformed, as shown among Fox News viewers during the Iraq war (46).

Although individuals may indeed engage in selective exposure in some circumstances, recent research suggests that this is not widespread in individuals’ everyday news consumption. A study tracking individuals’ online news use over time, for example, found no evidence of partisan selective exposure, instead concluding that the “online political news audience tends to overwhelmingly congregate within a handful of popular, brand-name news sites… [and that] all sites in the sample, including the more obscure, more partisan political news outlets, attract ideologically diverse audiences in proportion with the overall online audience” (47).

Still, as decades of research on motivated reasoning in political science (48), science communication (49), and other fields have shown, even when facts are not filtered out by selective exposure or “filter bubbles,” they can be interpreted in very different ways by different audiences. Even if ideologically diverse audiences are exposed to the same content, as the above study indicates, they may not do so for the same reasons or with the same outcomes (e.g., some Democrats may be “hate-reading” news on Breitbart News Network). Individuals, in other words, engage in goal-directed processing of new information to protect preexisting values, beliefs, and ideologies (50). When such directional goals influence reasoning processes, individuals are prone to “biased assimilation,” which is characterized by confirmation and disconfirmation bias, or the parallel tendencies to privilege information that is consistent with one’s predispositions and to discredit information that seems contradictory (51). As with selective exposure, motivated reasoning can contribute to an individual becoming misinformed, and it can occur not only in political contexts but also when individuals process information about science and emerging technologies (52–54).

Evidence also suggests that motivated reasoning is most likely to occur among individuals who have the most sophistication and knowledge of the topic at hand, further polarizing views among different publics (55). Unfortunately, simply providing individuals with corrective, factual information is not guaranteed to fix misperceptions, as “strongly entrenched beliefs” are often likely to “survive the addition of non-supportive evidence” (51). In fact, there is some evidence to suggest that attempts to correct misinformation about both political and scientific topics among individuals with the most strongly held beliefs can instead backfire, entrenching them further in their false views (56). However, some recent experimental work assessing people’s beliefs in eight ideologically polarized issues (none of which were science-related) has not been able to replicate this “backfire effect.” One interpretation for this is that individuals feel no need to “counter-argue” factual corrections (as the backfire effect suggests) because their ideological commitments may have more to do with affect than with evidence-based reasoning (57).

The role of emotion.

This brings us to discussions of the influence of affect in motivated reasoning processes. There is some evidence that a person’s emotional state can shape the accuracy of his or her beliefs. In recent experimental work, angry partisans who saw uncorrected political misinformation from their own party held less accurate beliefs than emotionally neutral partisans, raising concern that anger can facilitate belief in falsehoods, which might be “especially troubling given that anger also depresses information seeking and increases selective exposure” (58). However, when misinformation from an angry person’s in-party was coupled with a correction, there were no significant differences in belief accuracy between the two groups, suggesting that certain kinds of correctives may be effective despite emotional arousal.

Research also suggest that emotional states—especially anger—can interact with individuals’ ideologies and the information environment (e.g., the presence or absence of correctives, and in-group or out-group source cues) to influence people’s encounters with (mis)information, potentially exacerbating their belief in falsehoods and shaping how (mis)information is assimilated into their worldviews. Furthermore, recent research about the spread of falsehoods online has revealed that false rumors tend to elicit more feelings of surprise and disgust than the truth (24). Given that falsehoods were also shared more frequently, it is plausible that certain emotional states have greater power to inspire individuals to share information (24). Notably, individuals’ attraction to emotionally charged content is not limited to politics, and even when it comes to scientific discoveries, individuals are more inclined to spread information that has a greater emotional impact (59).

Addressing misinformation at the individual level.

As argued earlier, individuals may not only lack the ability to recognize and evaluate misinformation but also the motivation. Of course, citizens are constrained in their choices and behaviors by the information environments and institutions that surround them and that often produce content that is intentionally designed to circumvent the motivations of even the most well-intentioned news consumer. Acknowledging this additional wrinkle, research has started to examine ways to reduce motivational influences on how facts are processed.

For example, one proposal to achieve this might be to structure information environments in a manner that encourages accountability, since individuals are more likely to engage in effortful attempts to understand multiple sides of an issue—including on scientific topics—when they expect that their views will be challenged by others (60). Furthermore, instead of showcasing factual correctives or “the other side” of an argument sourced from out-group members or even an algorithm, it may be especially beneficial to source such content from like-minded others (e.g., “co-partisans”), whose arguments may be evaluated as more convincing (4). There may also be some value in disincentivizing expressions of partisan anger and outrage so they cannot be leveraged by disinformation campaigns to exacerbate biased assimilation of information (61).

Scientists.

Some scholars have argued that decreasing public and policy support for science, among other factors, has created new incentives for scientific organizations to use mass media and other public-facing channels to promote particular findings or areas of research. Media, in turn, rely on celebrity scientists as resources for newsworthy accounts of breakthrough science (49). This has engendered concerns that hype and overclaims in press releases and other scientific communication can lead to misperceptions among nonexpert audiences about the true potential of emerging areas of science, especially if science is unable to deliver on early claims of cures for diseases, for instance.

While there is limited empirical work on the effects of hype or overclaims on misperceptions about science being wrong or producing contradictory findings (84), survey data suggests that these kinds of concerns are not completely unfounded. In national surveys, one in four Americans (27%) think that it is a “big problem” and almost half of Americans (47%) think it is at least a “small problem” that “[s]cience researchers overstate the implications of their research.” Only one in four (24%) see no problem. In other words, while levels of confidence in the scientific community remain high (11), science may run the risk of undermining its position in society in the long term if it does not navigate this area of public communication carefully and responsibly.

Media.

While the scientific community continues to enjoy a great deal of trust, public confidence in the press has declined substantively over past decades. In surveys, only 1 in 10 Americans (8%) express a “great deal of confidence” in the press (8). Similarly, three in four Americans (73%) think that “[t]he biggest problem with news about scientific research findings is the way news reporters cover it” (12). Without a doubt, these poll numbers do not adequately capture all nuances of public trust in science or journalism as institutions (85). Even as imperfect indicators, however, they do not bode well for media’s ability to better inform various publics about science or to even counter misinformation among publics that are particularly critical of science.

Part of the problem is related to a well-documented decline in science journalism and “the trend among many media organizations to no longer use (full-time) science journalists” (11). As a result, coverage of scientific issues has often become the responsibility of political reporters, business writers, and journalists in other nonscientific beats (86). Paired with an increasingly polarized political environment, this has also promoted what some have called “false balance” (87). The term refers to reporting—often by nonscience journalists—that puts established scientific consensus around issues like genetically modified foods or climate change on equal footing with nonfactual claims by think tanks or interest groups for the sake of “showing both sides.”

Ironically, legacy media might also inadvertently be promoting their own demise by repeatedly and increasingly raising the specter of “fake news,” which myriad political actors have framed strategically as a disease of traditional media specifically to discredit them. Indeed, a search on Lexis Nexis for appearances of “fake news” in newspaper coverage in the United States and globally (Fig. 1) reveals a clear increase in the yearly frequency with which newspapers have uttered this specific phrase and have arguably made more familiar—and therefore more believable—its false connotations.

This is not particularly surprising. Media are in the business of reporting on political discourse, including potential accusations of bias in their own coverage. Recent research, however, also suggests that frequent discussion of elite discourse about fake news in news media may affect whether individuals trust news media (88). As a result, media may inadvertently undermine public trust in their own work by providing a forum for accusations of bias and fake news put forth by political elites.

Consumers of science news.

A third group central to the dynamics surrounding misinformation about science are consumers of science news themselves. A growing body of research suggests that the comments left by readers on news websites are frequently uncivil in tone, and that uncivil commenters are less likely to use evidence than civil ones (89). Even when the evidence that is being presented in user comments is held constant, however, uncivil comments following a science news article can promote perceptions among readers that the article itself is biased and can polarize readers based on their initial attitudes toward the technology that is being discussed (90, 91).

In other words, the nature of interactions among science news consumers can overshadow the quality of the information that is presented in an article or discussed in associated comments, and it can thus influence whether the information itself is seen as credible enough to be integrated into belief systems. Research is beginning to understand the processes behind these mechanisms as well as potential solutions (92), but there is still a very limited research base on how news environments that provide readers with real-time cues on popularity or partisan agreement (such as user comments, likes, or retweets) ultimately influence information processing and knowledge intake.

Systems approach.

The Academies’ Communicating Science Effectively report highlighted the need for a systems approach to the study of science communication—that is, the need for research that treats science communication as a multilevel problem. This suggestion acknowledges the fact that misinformation among individual citizens does not occur in a vacuum. Instead, individual misperceptions emerge in group-level processes as part of social networks, and they are embedded in and shaped by societal dynamics (such as political events or campaigns), as discussed above.

As a result, both researchers and practitioners of science communication have struggled to identify mechanisms for better tackling misinformation—especially in highly contentious areas of science—that generalize across issues and contexts. Particularly, there is limited research that examines how mechanisms for effective science communication that were originally studied in laboratory-experimental settings perform in competitive communication settings—for instance, when individuals are faced with competing messages from interest groups or political actors.

Systematic research on emerging media environments.

Calls for a systems approach also highlight a second area where research is needed, as outlined in the Communicating Science Effectively report: research on emerging media environments. There is new urgency to this problem, in light of findings that link use of social media to, for example, lower levels of political knowledge (94), as well as recent work suggesting that misinformation travels faster in social network than true facts (24).

One participant in the first Sackler Colloquium on the Science of Science Communication already highlighted the challenges for both researchers and practitioners resulting from a lack of a systematic corpus of work in this area: “[E]mpirical research examining specifically online science communication processes and outcomes is still scant … [and] many of the “best practices” of online science communication currently exchanged among practitioners are based on experiential evidence rather than on an empirical understanding” (95). Scholars have diagnosed similar research deficits related to misinformation and fake news in online communication environments. Their call to action mirrors systems approach thinking and calls for multidisciplinary efforts to empirically understand “the vulnerabilities of individuals, institutions, and society” to fake news disseminated through ubiquitous online channels (25).

Underserved audiences.

A last area with a dearth of systematic empirical work are mechanisms to reach audiences that are often underserved by traditional channels for science communication. Newspapers, science television, or even science museums, for instance, tend to reach more educated and higher-income audiences (11). Furthermore, from work dating back to the 1970s on widening knowledge gaps, we also know that citizens with higher socioeconomic status (levels of education and income) are able to learn more efficiently from new information than their peers. As a result, quality scientific information is not only more likely to reach more educated and higher-income audiences, but, when it does, the ability of citizens with higher socioeconomic status to process new information more efficiently can further widen existing gaps between the already information-rich and the information-poor.

Our inability to reach all segments of the population equally well with high-quality scientific information is particularly troubling, given that the need for antidotes to misinformation might be particularly pronounced among certain groups of the public. One indicator of this problem is the educational and income gaps related to seeing fake news as a problem in the first place. In a 2016 Pew survey after the US presidential election (28), almost three in four Americans (73%) with a household income over $75,000 thought that completely made-up news caused “a great deal” of confusion. Comparatively, fewer than three in five (58%) respondents with a household income of $30,000 shared that concern. However, socioeconomic gaps also emerge with respect to people’s confidence in their own ability to spot fake news in the first place. A 2018 US survey commissioned by The Economist asked respondents about their ability to distinguish real and fake news (96); 83% of respondents with a family income of at least $100,000 felt “very confident” or “somewhat confident” that they could “tell real news from fake news.” Among respondents with an income of less than $50,000, that number dropped to 63%.

Those self-assessments might be overly optimistic across all segments of the population (27). Nonetheless, gaps in perceptions of fake news as a problem and gaps in the ability to meaningfully access corrective information across groups with different socioeconomic status are troubling based on normative democratic standards. Indeed, they are especially worrisome as new technologies like CRISPR or artificial intelligence (AI) raise political and societal concerns that are of particular relevance to audiences who traditionally have been hard to reach for science communicators. For example, applications of AI for self-driving cars and other forms of automated services are raising concerns about changing labor markets that might disproportionately affect blue-collar workers. Similarly, expensive therapies based on new genome editing techniques raise questions of equitable access to treatments regardless of patients’ income or social status. All of these concerns will require political debates based on the best available scientific evidence. Those debates will have less-than-ideal outcomes if groups who will be significantly affected by these technologies are systematically uninformed or misinformed. Harnessing the science of science communication to ensure that socioeconomic disparities do not impact access to the best available scientific information for policy choices will therefore be crucial for the responsible development of these emerging technologies and new technologies in the future.