Skepticism toward science and scientists

In order to understand how scientists are evaluated, it is insightful to first turn to public attitudes toward science in general. Much existing research on science skepticism has focused on individual differences in worldviews and ideologies. Low faith in science and low willingness to support science (through the allocation of monetary resources) were found to be best predicted by belief systems ranging from religious orthodoxy to spirituality, thereby providing further support for the well-established oppositional relation between science and religion [11, 12]. Political conservatism is another key predictor of science skepticism, in particular toward climate science [11, 13, 14]. Additionally, many scientific debates, for example about genetically modified organisms (GMO), speak to people’s moral views about the world. In particular, concerns about purity and naturalness seem to have the potential to interfere with evaluations of scientific evidence if that evidence poses a threat to the individual’s moral values [11, 15]. In line with that finding, people have been found to selectively put their trust in scientific findings that are in line with their moral values [16]. Likewise, the degree to which people deemed a scientific hypothesis to be morally offensive negatively predicts the perceived credibility and acceptance of that hypothesis [17]. Moral values also play a role in research on specific domains of science skepticism, with data suggesting that both vaccine skepticism [18] and GMO resistance [11, 15] are at least partially driven by disgust and concerns about moral purity.

There is also evidence suggesting that science distrust is motivated by moral concerns about scientists. Rutjens and Heine [19] systematically investigated whether scientists are perceived as immoral across ten studies. Utilizing Moral Foundations Theory [20] as a framework, they found that participants perceived scientists as likely to commit breaches against binding moral foundations (i.e., loyalty, authority, purity), particularly against purity norms, yet not against individualizing foundations (i.e., care, fairness). Moreover, participants judged scientists to prioritize knowledge gain over morality, to lack human emotions, and even to be potentially dangerous. These findings are in line with other work on public perceptions of scientists, such as analyses of the image of scientists in Western literature and film. Tintori [21], for example, concluded that scientists are often portrayed as “isolated in their ivory tower, focused on their work, crazy, evil and dangerous” (p. 18). Others have suggested that scientists are perceived as having an unhealthy ambition in the pursuit of (forbidden) knowledge, which makes them potentially mad and dangerous [9, 22, 23]. Particularly this last idea is also reflected in survey data, which indicated that close to two thirds of EU citizens believe science to sometimes interfere with people’s sense of ethics, with half even believing scientists to “have a power that makes them dangerous” [24, see also 25]. These concerns about the morality of scientists are especially striking because other research finds that scientists are perceived as smart, highly competent, eminently respected, trusted, and even liked [26, 27]. In short, people seem to have quite ambivalent attitudes towards scientists.

Potential sources of contradictory findings

These opposing attitudes suggest there are hidden complexities in how scientists are evaluated that require further exploration. Complexities may lie both in how morality is conceptualized and operationalized, and in how people imagine what constitutes a scientist. Regarding morality, most previous studies have focused on Moral Foundation Theory’s categorization of the five moral foundations of care, fairness, loyalty, authority, and purity [20]. Yet, other theories of morality have employed the Stereotype Content Model (SCM), which proposes that people assess others along two primary dimensions of warmth and competence [28], by conceptualizing morality as one of two components of warmth (the other being sociability [29, 30]). Applying the SCM to scientists has revealed that scientists tend to be perceived as high in competence, but low in warmth [27]. Yet, an analysis that differentiates scientists’ perceived warmth into their perceived morality and sociability has yet to be conducted. Given the ambivalence in people’s views of scientists, distinguishing between the two components of warmth may be illuminating. Leach and colleagues [31] for example, applied this distinction in a study on anti-Semitic attitudes, showing that participants from the Russian Federation tended to perceive Jewish people as high in sociability but low in morality. Had only the overarching factor of warmth been examined, this important moral dimension of stereotypical perceptions of Jewish people amongst Russian participants would have likely been missed.

Second, the ambivalence that people hold towards scientists may be a product of how scientists have been defined or framed in the respective studies. Existing work has focused on perceptions of the “typical scientist” or of the broad group of scientists in general [19]. This approach may have neglected any nuances between different types of scientists, and it is quite likely that participants brought to mind quite different prototypes of scientists, thus adding much noise to the data. The present research sought to correct the limitations of past research by including different measures of morality and by asking people about different types of scientists, including self-generated examples.

Overview of studies

We conducted two studies aimed at investigating stereotypical perceptions of scientists, with a particular focus on perceptions of morality. Study 1 (conducted in March 2018) built upon insights from previous work that revealed associations between scientists and immorality [19], yet made use of more focused and well-established measures and materials. More specifically, we included two different measures of morality and employed different framings of scientists. In Study 2 (conducted in December 2018), we sought to replicate Study 1’s findings in another sample, using other control groups, a different design (between-subjects instead of within-subjects), as well as a different framing of scientists (a bottom-up approach of self-generated examples). In addition, Study 2 also explored the relationship between science skeptical attitudes and their relationship with perceived skeptical attitudes held by scientists as well as perceptions of scientists. All measures, manipulations, and data exclusions are reported.

Method

Discussion

The results from Study 1 replicate previous findings on social perception, suggesting that scientists tend to be perceived as highly competent [27]. With regards to warmth, the findings are mixed: scientists seem to be perceived as low in sociability yet high in morality, which stresses the importance of differentiating between these two components. Whereas the finding on sociability is in line with extant research that found scientists to be low in warmth [27], the finding on morality seems to contradict previous studies which identified associations between scientists and immorality [19]. Yet, different measures were employed to assess morality in the respective studies. It appears that scientists score comparatively low on the (binding) moral foundations [19], however, are generally perceived as moral when assessing morality along with sociability as a component of warmth, following the three-dimensional approach to stereotype content. This approach describes morality as consisting of honesty, sincerity, and trustworthiness–a definition that aligns closely with only one of the moral foundations, namely the individualizing foundation of justice/fairness (Note that sociability does not seem to map onto any of the moral foundations; although one could make an argument for sociability being somewhat related to either the harm/care or the loyalty foundation.). Hence, there is convergence across these studies–scientists are perceived as relatively immoral only when morality is conceptualized in terms consistent with the binding foundations. Within-comparisons for the university-affiliated and industry-affiliated scientists, for whom both measures of morality were obtained, support this idea: both groups scored higher on morality assessed as one of three dimensions of stereotype content, as compared to the moral (particularly binding) foundations. Thus, these findings stress the importance of incorporating different measures to assess the construct of morality.

Second, Study 1 tested different framings of scientists and found that perceptions of scientists depend on the type of the scientist in question. The scientist framed to work at a university was overall perceived more positively (in terms of sociability, competence, and morality) than the typical scientist. The scientist framed to work in the pharmaceutical industry, however, was perceived as less moral than the typical scientist, yet equal in terms of sociability and competence. Thus, perceptions of the typical scientist were overall closer to those of the industry-scientist as compared to the university-scientist. It is important to note that our example of industry scientists—scientists working in the pharmaceutical industry—might not generalize to other domains of industry science. However, the main goal of this aspect of the study was to show how different framings of scientists may lead to different social evaluations. In this light, it is also worth pointing out that the study was conducted before the COVID-19 pandemic.

This may indicate that people conceptualize scientists more as commercially working people who pursue economic benefits than employees of the public sector pursuing knowledge. Support for this differentiation comes from the finding that university-scientists were perceived as more curious than industry-scientists, while industry-scientists were perceived as more eager to satisfy their desire than university-scientists. These findings are consistent with findings by McCright and colleagues [34], who showed that attitudes toward science differed for what they coined impact-science (i.e., science that identifies environmental and public health impacts of economic production) as opposed to production-science (i.e., science that provides new inventions or innovations for economic production). Our descriptions of the university- and the industry-scientists may be playing at a similar differentiation, with university-scientists more likely conducting impact-science and industry-scientists more likely performing production-science.

Lastly, Study 1 finds that scientists are perceived as more moral than atheists (but not religious people) and less social than religious people (but not atheists). However, it is unclear whether these patterns are driven more by people’s views towards scientists or towards atheists and religious people, particularly given previous work suggesting people hold strong attitudes–negative and positive, respectively–toward atheists and religious people [35, 36]. To disentangle these effects, Study 2 included different—more neutral—control groups. Moreover, Study 2 moved beyond classical measures of social perception by also asking participants to make assessments about scientists’ and other groups’ attitudes toward controversially debated topics of science.

Method

Results

Analytic strategy.

We started by generating scales for binding moral foundations (loyalty, authority, and purity) and individualizing moral foundations (harm, justice). We also created a scale that reflected general agreement with the four science topics: namely, humans cause CO₂ emissions, vaccinations cause autism (reverse-scored), GMOs are safe, and humans have developed through evolution. Higher scores reflected opinions more in line with scientific consensus. The internal reliabilities for each of these scales within each of the targets varied from .53 to .73. Internal reliabilities for each group separately as well as zero-order correlations for all variables can be found in the S3 and S4 Tables.

To test for differences in ratings between the typical scientist, novelist, citizen, and self, as well as the first-to-mind prominent scientist and first-to-mind prominent novelist, we took the same approach as in Study 2. First, we conducted univariate ANOVAs, which were followed up by Tukey HSD tests (see Table 2). Again, this procedure was run for both the individual items, as well as the scale totals. Then, we used linear regression to test the extent to which participants’ agreement with the controversial science topics was predictive of a) their perceived agreement of the typical scientist with those topics as well as b) their perceived morality, warmth, and competence of the typical scientists. Furthermore, we repeated the same regression models with a different predictor, namely a scientist/novelist-knowledge score that reflected the extent to which participants correctly listed three examples of prominent scientists or novelists, which could range from 0 to 6: For each of the three provided examples, participants either scored 0 (no scientist named), 1 (person named is science-affiliated, but prominent for other skills e.g., business person Elon Musk), or 2 (person named is clearly scientist, e.g., Stephen Hawking). The same logic was applied to the novelist condition. Additionally, we tested the extent to which this knowledge score was predictive of the participant’s own agreement with the controversially debated science topics.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Means and standard deviations for all groups and results of ANOVA and Tukey HSD tests.

https://doi.org/10.1371/journal.pone.0274379.t002

Data analysis

All univariate ANOVAs were significant and are reported in Table 2 along with means and standard deviations for each item and scale. Tukey HSD tests showed that the typical scientist was perceived as more competent compared to both the typical novelist (p = .007, d = 0.506) and typical citizen (p < .001, d = 1.237). In terms of warmth, the results were mixed. Again, scientists were perceived as less sociable than both control groups (ps < .001, d_novelist = 0.669, d_citizen = 0.812). However, in terms of morality, a more complex picture emerges. Although scientists scored (marginally) higher on morality as a dimension of stereotype content compared to novelists (p = .090, d = 0.328) and citizens (p = .020, d = 0.319), they scored lower on almost all moral foundations. On the scale level, the typical scientist was perceived as less motivated to abide to individualizing moral foundations compared to the typical novelist (p < .001, d = 0.604) and citizen (p < .001, d = 0.549)–a finding that holds for both individual level foundations of harm/care (ps < .001, d_novelist = 0.619, d_citizen = 0.656) and justice/fairness (p_novelist = .008, d_novelist = 0.457; p_citizen = .010, d_citizen = 0.359). Similarly, the typical scientist was rated lower on binding moral foundations compared to the typical citizen (p < .001, d = 1.024), yet not compared to the typical novelist (p = .993, d = 0.084). For comparisons between the typical scientist and typical citizen, this pattern also holds on the individual level for all three binding moral foundations of loyalty (p < .001, d = 0.772), authority (p < .001, d = 0.742), and purity (p < .001, d = 0.738). The typical scientist’s agreement with controversially debated science topics was judged to be higher on all four topics than the typical novelist’s or citizen’s agreement, both on the scale (p_novelist < .001, d_novelist = 1.118; p_citizen < .001, d_citizen = 1.337) as well as on the individual item level (all ps < .001, ds > 0.546). Additionally, participants also completed ratings for themselves. In line with research on self-enhancement [37], participants generally rated themselves as more positive compared to the other groups, especially in terms of morality and warmth. Specifically, they perceived themselves to be more moral (ps < .001, ds > 0.667) and warmer (ps < .001, ds > 0.932) than all other groups. In terms of competence, they perceived themselves as more competent than the typical citizen (p < .001, d = 0.968) and half-way between the highly competent scientist (p = .221, d = 0.243) and the somewhat competent novelist (p = .477, d = 0.213). With regard to the moral foundations, participants rated themselves as more motivated to comply with individualizing moral foundations than all other groups (ps < .001, ds > 0.667). However, for binding moral foundations, participants self-evaluations scored closer to the lowest-scoring typical scientist (p = .691, d = 0.146) then to the highest-scoring typical citizen (p < .001, d = 0.672).

In the second part of Study 2, we found that the most frequently mentioned prominent scientists were all physicists: Albert Einstein, followed by Stephen Hawking, and Marie Curie. In comparison to the novelist condition, participants in the scientist condition generated fewer unique examples (69 scientists compared to 148 novelists) and were also less like to generate correct examples: only 53.91% of the participants were able to list three correct examples of prominent scientists, compared to 88.23% in the novelist condition (although this difference almost vanishes when only looking at the first-to-mind example: 87.50% correct for scientists, 91.60% for novelists).

In general, for both the scientist and novelist condition, the prominent examples that participants generated were rated more positively than their typical counterparts. Specifically, in comparison to the typical scientist, first-to-mind scientists were rated as significantly more sociable (p < .001, d = 0.594) and as more motivated to follow the moral foundation of harm/care (p < .001, d = 0.679). In terms of controversially debated science topics, the first-to-mind scientists tended to be somewhat more skeptical (i.e., less in line with scientific evidence) than typical scientists (p = .023, d = 0.376; especially on the human CO₂ statement: p = .021, d = 0.359; and GMO safety statements: p = .023, d = 0.414), thus rendering them more similar in ratings to typical novelists, citizens, and the self. When making comparisons between perceptions of the typical novelist and the first-to-mind novelist, fewer differences surfaced as compared to the scientist condition.

Further, we explored whether the degree to which participants believed scientists to agree with the statements on controversially debated science topics could be predicted by their own agreement with these topics. Indeed, for each topic, participants’ own agreement was highly predictive of their perceived agreement of scientists with that statement (human CO₂: b = .276, t(125) = 3.514, p < .001; vaccinations: b = .477, t(125) = 5.835, p < .001; GMO safety: b = .362, t(125) = 3.604, p < .001; evolution: b = .394, t(125) = 4.068, p < .001). Moreover, participants’ agreement with the science statements was also predictive of their perception of scientists in terms of morality, sociability, and competence. The extent to which participants agreed with the theory of evolution positively predicted their perceived morality of scientists, b = .258, t(122) = 2.695, p = .008. The extent to which participants believed in anthropogenic climate change predicted their perceived sociability of scientists, b = .200, t(122) = 2.008, p = .047. And last, participants’ perceptions of scientists’ competence was predicted by their agreement with anthropogenic climate change (b = .281, t(122) = 2.954, p = .004), the extent to which they disagreed with vaccinations causing autism (b = .220, t(122) = 2.315, p = .022) and the extent to which they did not agree with GMOs being safe (b = -.220, t(122) = -2.211, p = .029). (The predictive relations between participants’ agreement with controversially debated science topics and their perceived agreement of scientists with those topics as well as their perceived morality, warmth and competence of scientists also hold on the scale level, that is, when combining participants’ agreement with the four statements on controversially debated science topics into one score.)

Next, we tested the same models with a different predictor, namely the extent to which participants were able to correctly name three prominent examples of scientists. On average, participants achieved a scientist-knowledge score for the three prominent examples of 5.35 (out of 6, SD = 0.81). This scientist-knowledge score positively predicted participants’ perceived agreement of the typical scientist with the CO₂ statement (b = .269, t(126) = 3.137, p = .002) and of the conception that vaccinations do not cause autism (b = .228, t(126) = 2.624, p = .009). For the typical scientist’s perceived morality, no effects of the scientist-knowledge score were observed. Lastly, we also tested whether participants’ scientist-knowledge score was predictive of their own agreement with the statements on controversially debated science topics. Indeed, their scientist-knowledge score positively predicted participants’ agreement with the conception that human emissions cause CO₂ (b = .349, t(125) = 4.166, p < .001) as well as their agreement with the theory of evolution (b = .197, t(125) = 2.241, p = .027).

Discussion

To a large extent, Study 2 replicates the findings of Study 1. As in Study 1, participants perceived scientists as less sociable (here, warm) and more competent compared to control groups (i.e., typical novelist or citizen). In addition, Study 2 yielded mixed findings on morality, revealing that while scientists tend to be perceived as moral when assessing morality along the three-dimensional approach to stereotype content, they seem to be perceived as comparatively immoral in terms of their compliance with the five moral foundations. Interestingly, while Study 1, in line with previous research [19], found associations between scientists and immorality to only hold for binding moral foundations, the results of Study 2 suggest that scientists also score significantly lower on individualizing moral foundations compared to the control groups. In fact, while within-comparisons show that scientists are indeed still perceived as less moral in terms of binding as opposed to individualizing moral foundations, similar comparisons within the control groups indicate that this difference is actually smaller for scientists as compared to novelists, typical citizens, or even the self. This results in larger between-group differences for individualizing as opposed to binding moral foundations between scientists and control groups, which–contrary to previous findings–may suggest that it is the relative neglect of individualizing foundations which accounts for more negative attitudes toward scientists as compared to control groups. This may especially be the case given that binding moral foundations seem to be less valued than individualizing foundations in general, as suggested by a) the lack of self-enhancement in the self-ratings as compared to ratings on the typical citizen, b) low correlations between binding moral foundations and positive dimensions such as sociability and competence (see Supplementary Materials), and c) Study 1’s positively perceived university-scientists only differing from the generally negatively perceived industry-scientist in terms of individualizing, but not binding moral foundations. Thus, even relatively small breaches against individualizing moral foundations may lead to substantial negative perceptions of scientists.

Similar to Study 1, Study 2 also investigated the effect that scientist framing has on stereotypical perceptions of scientists. In Study 2, participants not only made judgments about the typical scientist, but also about the first prominent scientist that came to their mind. Those first-to-mind scientists were overall rated more positively than the typical scientist, especially in terms of sociability and individualizing moral foundations (that is, where perceptions of typical scientists are most negative); a finding which also emerged comparing the first-to-mind and typical novelists. This could reflect a general effect of rating prominent examples instead of the prototype of an occupational group, as prominent examples are usually prominent for their positive impact on the world. Yet, it is interesting to note that fewer differences surfaced in the novelist condition as opposed to the scientist condition, indicating that there must be a scientist-specific effect. Combining this with the finding that participants had a harder time coming up with three correct examples of scientists as compared to novelists, and also mentioned fewer unique examples, our findings may suggest that participants’ attitudes toward the typical scientist are more heavily influenced by behavior that participants consider possible instead of what is probable (as low familiarity with scientists may not allow for an estimate of what is probable). In addition, attitudes toward typical scientists may be more heavily influenced by fictional scientist characters (e.g., cultural archetypes of evil scientists, such as Mary Shelley’s Dr. Frankenstein [38]), which–as we know from existing work on the image of scientists in pop-culture–are often portrayed as mad and dangerous, due to their unhealthy ambition in the pursuit of knowledge [9, 22, 23]. Empirical support for the idea that these cultural images of scientists may influence stereotypical perceptions of scientists comes from previous research showing that scientists are perceived as willing to prioritize knowledge gain over doing “the right thing” and thus are potentially more dangerous [19].

Lastly, Study 2 showed that the extent to which people are skeptical about science predicts their perceptions of scientists with regards to morality, warmth, competence, and the perceived standpoint of scientists on such controversies. Even more interesting though, the extent to which participants were able to correctly name three prominent scientists was also predictive of the perceived agreement of scientists as well as their own agreement with the controversial science topics. This suggests that an individual’s personal familiarity with scientists may be associated with their attitudes about science in general–an idea that resonates with recent suggestions that a person’s psychological distance to science (here, probably particularly perceived social distance) may impact their skepticism toward specific scientific debates or science in general [39] (see also [40, 41] for research on psychological distance and climate change denial).

Limitations

There are several limitations to the reported studies. First, as in any research studying attitudes towards specific groups, the choice of reference or comparison groups has an impact on the interpretation of results. No obvious comparison group exists for scientists, making the selection of appropriate control groups a difficult task. In the above studies, we tried to overcome this issue by comparing scientists with a variety of different groups, paying attention to selecting both generally positively perceived (religious people) as well as negatively perceived groups (atheists), in line with previous work [36]. Moreover, we also included comparisons with more neutral control groups (novelists, typical citizens), the self, and even comparisons within different types of scientists. Thereby, we aimed to capture a reliable picture of how scientists are perceived as opposed to how they compare to a specific group.

Second, a central finding in Study 1 was the divergent perceptions of university-affiliated as opposed to industry-affiliated scientists, with the latter generally perceived more negatively than the former. However, it is important to note that our Study 1 sample consisted of a convenience sample of university students. Therefore, it is possible that demand characteristics or mere exposure effects played a role in the more positive attitudes toward university-affiliated scientists. Although we addressed the limitation of using a convenience student sample by sampling through Amazon Mechanical Turk in Study 2, the specific finding on distinct perceptions of university- as opposed to industry-affiliated scientists warrants replication in a representative sample.

Third, in our research design, we did not specifically control for how perceptions of scientists may depend on the perceived gender of the scientist in question. In Study 1, only male personas were introduced through our stimulus material, thus preventing gender from acting as a confounding variable, yet at the same time neglecting it as a possible moderator. By using a bottom-up approach in which participants generated examples of scientists themselves, we moved beyond this limitation in Study 2. Yet, we found that most participants still thought of male as opposed to female scientists–only 8.5% of the generated examples were female. These findings are in line with years of results obtained with draw-a-scientist experiments [42, 43]. Thus, findings from Study 2 still largely reflect perceptions of male as opposed to female scientists–an issue that may have even been exaggerated by our unfortunate use of gender-biased language in one of the comparison groups (“typical countryman”). Thus, future research should play close attention to using gender-neutral language and systematically test how perceptions of scientists interact with gender perceptions.

Implications

Despite legitimate limitations of the present research, our findings have implications for a range of research lines. First, they hold important insights for morality research in general, in showing that the conceptualization and operationalization of morality has serious consequences for the findings one can expect to obtain–to the extent that one might report opposite findings depending on the morality measure in question. Existing research seems to often overlook this ambiguity in the construct of morality. Second, our findings inform general research on stereotyping by highlighting the importance of differentiating between the sociability and morality components in the warmth factor of the SCM. Until today, much research on stereotyping continues to neglect this differentiation of warmth [44–46], which has serious ramifications for research on groups that are perceived as high in morality and low in sociability—such as scientists—or the other way around.

Most importantly, our findings provide insights into attitudes toward scientists. First, they suggest that framing effects may be able to explain and account for some ambivalent or contradictory findings within the current body of literature, as even slight differences in framings of scientists between studies may have triggered divergent perceptions. Second, these effects may not only play a role in attitudes toward scientists, but also toward science in general. The type of science in question (production vs impact science) should thus be taken into consideration when assessing potential drivers behind science skepticism. And lastly, the observed framing effects have important implications for science communication. As we found people to hold more negative attitudes about scientists working in industry as opposed to scientists working in academia, it may be advisable to emphasize scientists’ university backgrounds when they communicate with the public. This may be particularly important whenever there is an association with pharmaceutical work (the example industry used in the present research). Highlighting work by university-affiliated researchers or actively stressing certain qualities in the industry scientists (e.g., pursuing knowledge instead of financial gain) may help to shape more positive attitudes toward the respective scientist. Given the observed relations between attitudes toward scientists and skepticism toward science in general, this may then help to decrease overall skepticism toward the integrity of the presented scientific findings–an objective of high importance in light of the problem of public repudiation of science.