Science and scientific progress have been essential to foster modernization and development (Merton, 1942). In today’s societies, however, much of this success depends on the public’s acceptance of and trust in scientific methods and findings. Science can only act as an independent source for evidence-informed policy-making and provide a common factual baseline for public discourse if it has widespread support from the public (Wilholt, 2013). The past decade has revealed a shift in public perception of science among certain subgroups and populations. The opposition to COVID-19 vaccinations (Lazarus et al., 2022; Sturgis et al., 2021) or the rejection of the existence of climate change (Dunlap, 2013; Herranen, 2023) are just two examples of this paradigm shift. If trust in science is waning, this could significantly affect the ability of humanity to address some of the most pressing issues of our times. This has spurred efforts to understand the origins of science skepticism globally, and particularly the political contexts that foster it (Lewandowsky et al., 2016; Rutjens et al., 2022; Jiang & Wan, 2023).

Some have long assumed that democracies provide environments that are most conducive to science, where enlightenment and education should lead to higher levels of trust in science (Edel, 1944; Schroeder, 2021; Sigerist, 1938). Recent developments, however, have shown that this no longer holds. Open and public debate—a fundamental aspect of democracy—also exposes scientists and scientific findings to challenges from competing opinions and interests, which can contribute to skepticism toward scientific work at large (Lewandowsky et al., 2016; Mede & Schäfer, 2020). Comparative studies like those by Gauchat (2012) and Makarovs and Achterberg (2018) offer limited empirical support for either assumption. In fact, some cross-national surveys indicate that average trust levels in democracies are only slightly higher than in autocratic countries (see SI section 1).

These surveys, however, are problematic for several reasons. They assume that respondents across very different cultural and political contexts have a shared understanding of “science”, something we cannot take for granted. Furthermore, existing surveys do not distinguish between different disciplines, which could be perceived very differently under different political regimes. Since we do not know whether respondents in different political regimes have similar conceptions of science, we do not know whether their ratings are actually comparable. The seeming equivalence of trust in science across countries and regimes may be masking different underlying perceptions of science and the people involved in it.

In our study, we address these issues by conducting a preregistered vignette experiment in ten countries that vary in political regimes, from highly autocratic to fully democratic. Respondents in ten different countries rate fictitious scientific findings, in which we vary the characterization of the scientists and their findings as described above. We present the findings in the form of a short headline summary that would appear in everyday media. Our sample includes China, Brazil, India, Mexico, Singapore, Poland, South Africa, Turkey, Germany and the United States. For each vignette, we maintain the distinction between scientists and science. Respondents provide six different ratings: three regarding the scientist’s characteristics (perceived competence, honesty, and responsibility) and three regarding the science itself (general trustworthiness, suitability for public funding, and policy influence).

In addition to the core experiment, we examine key survey variables to better contextualize our findings. While our primary focus is on how the characteristics of scientists and science affect trust across different countries and political regimes, these differences may at least partly reflect cross-national variation in education and media consumption. For example, political regime type does not directly moderate the relationship between education and trust in science, however, in liberal democracies, more educated people have greater trust in science, while the opposite holds in authoritarian regimes where less educated people trust science the most. The exposure to science through deliberate media consumption plays out similarly; the gap between low and high consumers is equally pronounced across political regimes. Lastly, previous research has shown heterogeneity in scientific skepticism among people’s political orientations. In the United States, for example, vaccine skepticism and COVID-19 misinformation were most pronounced among conservative individuals (McLamore et al., 2022). Consequently, we explore education, political orientation, and science media consumption as covariates in our study, although they are not experimentally manipulated.

The broad definition of science is not limited to a singular, monolithic concept. It relates not only to the scientific method, which builds knowledge through systematic and rigorous principles of inquiry but also to the actors engaged in producing scientific knowledge and institutions supporting and shaping how science is practiced, implemented, and impacts society (Merton, 1942; Nagel, 1961). Substantial heterogeneity in public understanding of science is introduced through institutional systems of science and science education, and individual experiences within these environments (Bromme & Goldman, 2014). Political interests and ideological motivations interact with these systems, resulting in significant differences in beliefs, perceptions, and attitudes towards science across contexts (Guo et al., 2022; Osborne et al., 2003). The influence of scientific culture is evident in public representations of science, which have evolved to encompass specialized knowledge in numerous scientific domains (Bauer et al., 2006; Bromme & Goldman, 2014; Ge et al., 2021; Godin & Gingras, 2000).

Following Schmitter and Karl (1991, p. 4), we define political regime as the institutionalized norms, rules, and structures that determine the organization of a government, encompassing how power is obtained, exercised, and transferred within the framework of political interactions and decision-making processes. For this study, and in line with most research in political science, we examine variation in political regimes as regards their level of democracy, broadly defined as respect for liberal values such as “constitutionally-protected civil liberties, strong rule of law, and effective checks and balances that limit the use of executive power” (Coppedge et al., 2024, p. 4). Education and cultural socialization of individuals are influenced by the norms and values of the political regimes in which they live. This long-term exposure shapes attitudes, beliefs, and behaviors, molding an individual’s worldview and perception of scientific and political authority (Lewandowsky & Oberauer, 2021). With few exceptions, democratic values within countries evolve slowly over time (Hoffmann-Lange, 2024). Consequently, we only focus on the level of democracy in the contemporary political regime during the year of our vignette experiment (i.e., 2023).

Governments set their own strategic objectives and resource allocation priorities, shaping education policies and public discourse around science systems, particularly scientists, institutions, and research outputs. In authoritarian regimes, indoctrination and repression mean that science serves state interests, while democratic regimes typically prioritize scientific freedom, critical thinking, and international cooperation. These differences reflect broader societal values and governance structures that shape attitudes toward scientific inquiry and its role in society (Edel, 1944; Josephson, 2005; Kitcher, 2003). Ignoring disparities in political systems overlooks scientific culture, rendering an incomplete and inaccurate picture of public trust in science and scientists.

We disaggregate the concept of science along several characteristics. Most importantly, we provide a more fine-grained distinction between perceptions based on scientists as the persons involved in science, and perceptions of science itself, namely the research topics and the outcomes that science generates to foster knowledge and inform policy. As shown by Achterberg et al. (2017), this distinction is important because people may place great trust in scientific methods and principles, but simultaneously distrust the people and institutions applying these principles. In democracies, science skepticism has been shown to be partially driven by anti-establishment and anti-elitism sentiment that rejects scientific knowledge because it originates from technocratic or academic “elites” (Mede & Schäfer, 2020). In other words, this skepticism arises from suspicions about scientists involved in knowledge production.

On the other hand, popular concerns about science could also be affected by varying perceptions of different academic disciplines (Altenmüller et al., 2024) and, in particular, the perceived social and political implications of their research. By separating the perceptions of scientists from those of their research, it becomes possible to focus more clearly on the evidence, data, and methodologies underpinning scientific knowledge, minimizing personal biases or stereotypes (Finson, 2002). This distinction allows for a more precise evaluation of science itself, as perceptions can be based on the merits of scientific output rather than preconceived notions about scientists (Dahlstrom, 2014; Jones et al., 2000).

Comparative studies by Allum et al. (2008) and Bauer et al. (2011) have highlighted significant differences in public understanding and knowledge of science across countries but without accounting for the influence of their political systems. While there is no direct evidence comparing determinants of trust in science in democracies and non-democracies (except the recent study by Jiang and Wan (2023)), existing country-specific studies in scientific domains like climate change, evolution, genetics, and vaccinations suggests that there may indeed be politically-motivated variation that remains woefully understudied (Bertoldo et al., 2019; WonPat-Borja et al., 2012; M. Zhang et al., 2022). Following this, we assume that cross-national variation in ruling ideologies, education systems, and societal values leads to different notions of the concept of science, which inform citizens’ beliefs and attitudes differently (Figueiredo et al., 2020; Ge et al., 2021).

Our approach contrasts with large cross-national observational data collections, such as the recent TISP Many Labs project (Cologna et al., 2024). We use a preregistered, cross-national survey experiment (N = 8441; see Methods), which allows us to examine the causal effects of specific characteristics of both scientists and science across countries with different political regimes. For the former, we focus on different personal characteristics of scientists that are likely to be perceived differently in different countries. By relying on random assignment of participants to different treatments, we also rule out that systematic effects of particular population sub-groups are driving the perceptions we measure (Raza et al., 2002).

Perceptions of scientists

Public perceptions of scientists are often stereotypically informed; scientists are viewed as competent and intelligent but also socially awkward, cold, and aloof (Altenmüller et al., 2024; Ferguson & Lezotte, 2020). Trustworthiness is a complex amalgamation of value judgments based on competence, integrity, and benevolence, and recent studies have found that self-disclosure of personal information can improve people’s perception of benevolence and integrity of disclosing scientists (Altenmüller et al., 2023; Fiske & Dupree, 2014; Schinske et al., 2016). However, such self-disclosure comes with a trade-off between being perceived as more competent or more “warm” and ultimately does not increase credibility or trust in their scientific research (Altenmüller et al., 2023). Given our interest in cross-national comparison, we examine how the public assessment of scientists depends on highly salient characteristics and direct engagement. Perhaps the most apparent, well-studied factors are the scientist’s gender, nationality, and level of public activism.

Research has consistently shown that specific stereotypes are ascribed to scientists based on their gender presentation, with female experts often perceived as less competent and less trustworthy than their male counterparts (Banchefsky et al., 2016; McKinnon & O’Connell, 2020; Ozer, 2023). This gender bias affects the evaluation of both scientific research and the credibility of scientists. For instance, Ozer (2023) found that female experts in politics were perceived as having less expertise than males with identical qualifications. McKinnon and O’Connell (2020) show that people are more likely to hold negative stereotypes about female scientists who publicly discuss their work. Incongruent beliefs about the traits of successful scientists and ideas about women being nurturing, warm, or passive can also lead to prejudice against female scientists (Banchefsky et al., 2016; Carli et al., 2016). Implicit stereotypes about women’s lower aptitude for STEM disciplines can lead to biased hiring decisions that impede their scientific career trajectories (Reuben et al., 2014). This suggests that gender biases have a significant impact on the representation of female scientists and people’s evaluation of their credibility and support for their scientific research. As gender equality varies considerably across countries, with autocratic countries often exhibiting greater discrepancies in the treatment of men and women (World Economic Forum, 2023), this bias may influence the perceptions of scientists across different political regimes.

We further examine how the nationality of the scientist—whether the scientist is a co-national or a U.S. citizen—affects public trustworthiness assessments. Research has shown that individuals tend to trust medical doctors and scientists from their own country more than those from other countries (Chavarria-Soley et al., 2021). According to social identity theory, in-group favoritism may stem from a sense of familiarity and shared cultural background, leading to a perception of greater reliability and understanding for co-ethnics or co-nationals, regardless of their actual qualifications or expertise (Jia & Luo, 2023; Luo & Jia, 2022). Luo and Jia (2022) find that this inclination is closely linked to nationalism; individuals who have greater trust in “national” science tend to be more politically aligned with their government. Therefore, we expect that people will trust scientists and scientific findings from their country of origin more than those from other nationalities. Since nationalist and nativist ideas are central to many autocratic governments (Panzano, 2024), we further expect that the scientist’s nationality may have a stronger impact on perceptions of trustworthiness in autocratic regimes.

Moreover, there is ongoing debate about the role of activism and political engagement in science. Some argue that scientists should remain neutral and avoid taking political stances, while others believe that activism is essential for promoting social change and addressing societal issues (Lupia, 2023; Motta, 2020). Research suggests that perceptions of scientists engaged in public activism vary along partisan lines in polarized democracies like the U.S. (Guenther et al., 2019; Motta, 2018). Scientists who publicly advocate for causes may be perceived as less trustworthy and less credible by conservative individuals who view their “liberal-activist stance” as a threat to the objectivity and neutrality of science (Bergan et al., 2022; Cofnas et al., 2018). However, other studies find that engaging in activism without advocating for specific policies is not detrimental to scientists’ credibility or objectivity and can improve public awareness and societal outcomes (Cologna et al., 2022; Kotcher et al., 2017; Parsons, 2016). Additionally, advocate scientists may be perceived as having greater expertise but only as trustworthy as laypersons on some topics (Geiger, 2022). In some countries, such as China, greater grassroots advocacy and fewer institutional accolades may actually boost the credibility of scientists (J. Y. Zhang, 2015). At the same time, the politicization of issues like climate change can contribute to misperceptions about the motivations of activist scientists, thereby negatively affecting public trust in certain countries (Cloud, 2020; Hmielowski et al., 2014) but not in others (Cologna et al., 2021). Additionally, activism may shape public perceptions of scientists differently in democracies with a strong tradition of public activism and formal protections for protest, than it does in autocratic regimes.

Perceptions of science

Regarding the science itself, we focus on two key dimensions. We assess how trustworthiness ratings differ between scientific disciplines and their typical research projects (Gligorić et al., 2022; Tranter, 2023). Based on prior research (Altenmüller et al., 2024), we assume that perceptions differ between what is colloquially referred to as “hard” sciences (STEM, engineering) and “soft” sciences (economics, social sciences), which should affect the level of trust in these fields. Growing evidence shows that people perceive significant differences between disciplines: Natural sciences are perceived as more authoritative and competent and benefit from greater support in policy contexts, while social sciences are perceived as more sociable and moral (Gligorić et al., 2024; O’Brien, 2013). Furthermore, many autocratic regimes do not freely tolerate various disciplines of the “soft” sciences as they may pose threats to the regime (S. Bergan et al., 2020; Kinzelbach et al., 2023). Instead, they place a stronger emphasis on “hard” science disciplines, which may further widen the differences in perceptions of disciplines within autocratic regimes compared to democracies.

Studies have shown that trust is perceived differently across various disciplines, such as sociology, psychology, computer science, and economics (Gauchat & Andrews, 2018; Scheitle & Guthrie, 2019). This may stem from preconceived notions about certain disciplines, or be an artifact of funding priorities or scientific culture within countries (Godin & Gingras, 2000; Scheitle & Guthrie, 2019). For instance, the replication crisis in the social and psychological sciences led to a crisis of confidence among scientists and, consequently, the public (Anvari & Lakens, 2018). Politically biased generalizations cause credibility issues for entire disciplines like sociology and nutrition science, highlighting the broader challenges of trust in the scientific community as a whole (Penders et al., 2017; Scheitle, 2018). Furthermore, public confidence in some domains has become increasingly politicized, with variations in the perceptions of knowledge and consensus among scientists in climate science and public health (Critchley, 2008; O’Brien, 2013; Scheitle & Guthrie, 2019).

The relationship between science and society is bidirectional, with society expecting science to address societal needs and expectations (Contessa, 2022). It comes as no surprise that the perceived impact of scientific research also influences public trust. Since people have differing expectations of research, their perceptions of what constitutes high-impact research also vary (Resnik, 2011). Some studies find that those who have higher interest and engagement with research on new technologies (e.g., nanotechnology) also perceive it to have greater societal benefit (Retzbach et al., 2011). High-impact research sees greater public consumption, and this public use drives greater funding towards high-impact research areas (Yin et al., 2022). Emphasizing the altruistic motives behind scientific research can enhance public trust as scientific endeavors are perceived to have greater societal benefit (Benson-Greenwald et al., 2023). In contrast, people consider commercially-oriented research much less trustworthy due to misapprehensions about the credibility of findings or doubts about manufactured consensus for results aligned with corporate interests (Pinto, 2020). Furthermore, we expect that differences in the perception of corruption across countries and political regimes also apply to ratings of trust in high-impact research (Alper, 2023).

In our experiment, we avoid highly polarized research topics, such as climate change, vaccinations, or COVID-19. This is for two main reasons: first, there is considerable research that focuses on public attitudes and trust in institutions in the face of global crises, and second, our focus on less polarized domains ensures greater experimental control as participants’ trust in scientists and science is not confounded with their issue attitudes. Instead, we focus on four specific disciplines as tangible examples within the big streams of research: materials science as a natural science discipline, genetics as a discipline of the life sciences, and economics and education studies as examples of social science disciplines. We also analyze to what extent the perception of science depends on its public impact, expecting trust to be higher as this impact increases.

Thus, we formulate the following hypotheses for this study:

H1 Respondents’ trust in the scientist, in the science presented, as well as their support for public funding and policy influence of the science, will be as follows: genetics > materials science > economics > education studies.

H2 Respondents’ trust in the scientist, in the science presented, as well as their support for public funding and policy influence of the science, will be lower for female scientists compared to male scientists.

H3 Respondents’ trust in the scientist, in the science presented, as well as their support for public funding and policy influence of the science, will be lower if scientists engage in activism.

H4 Respondents’ trust in the scientist, in the science presented, as well as their support for public funding and policy influence of the science, will be higher if the science affects many people.

H5 Respondents’ trust in the scientist, in the science presented, as well as their support for public funding and policy influence of the science, will be higher if the scientist shares their nationality with the respondent.

In line with our preregistration, we examine the impact of respondent characteristics—education level, science media consumption, and political orientation—on trust in science. This covariate analysis enables us to explore how these individual-level factors interact with the political regime in which an individual is socialized to shape perceptions of scientific authority.

The role of education in fostering trust and confidence is well-established. Individuals with higher levels of education are more likely to trust science and scientists (Achterberg et al., 2017; Allum et al., 2008; Alper et al., 2023; Bak, 2001; Pullman et al., 2019). Although societies with higher education levels generally exhibit more trust in science, within advanced societies, wider provision of mass education has not consistently resulted in increased trust, possibly due to rising politicization and anti-elitism (Gauchat, 2012). This discrepancy suggests that factors beyond education may be equally influential.

Research overwhelmingly shows that political ideology is an increasingly salient determinant of trust in certain types of scientists and science: liberals display greater trust than conservatives (Azevedo & Jost, 2021; Li & Qian, 2022), particularly on polarized issues (Drummond & Fischhoff, 2017; Kampourakis & McCain, 2019; McCright & Dunlap, 2011; Schrøder, 2022). Importantly, this dynamic does not persist to the same degree in non-democracies with different political alignments (McLamore et al., 2022).

Exposure to science, particularly direct exposure (e.g., looking up information on the internet) compared to indirect exposure (e.g., coming across information on social media newsfeeds), can influence perceptions about scientific issues and scientists themselves (Bromme & Goldman, 2014). When individuals engage with scientific information, they have the opportunity to form beliefs and develop attitudes about the trustworthiness of scientists and the credibility of scientific information, both of which are key determinants of trust in science (Fiske & Dupree, 2014; Wintterlin et al., 2022). The effect of exposure on trust (and vice versa) also largely depends on the quality and nature of the information received (Bromme et al., 2010; Rosman & Grösser, 2024). Accurate, well-explained, and unbiased scientific information tends to increase trust (Brewer & Ley, 2013; Wilson et al., 2017). Conversely, exposure to misinterpreted, sensationalized information, or pseudoscience can lead to distrust or skepticism (Lewandowsky et al., 2022; Scheufele & Krause, 2019).

The questionnaire consisted of two modules. The first module had two sections: (1) demographics and (2) science education, understanding, and exposure (pre-treatment). The second module contained the vignette experiment (Steiner et al., 2017). The main analysis relies only on attribute treatments and trust outcomes from the experiment (see Vignette Experiment). The secondary covariate analysis relies only on items drawn from the pre-treatment module (see Relevant Covariates). The research design for this study was preregistered and is available on OSF: https://osf.io/f4y7v.

Participants

This cross-national survey was fielded among adults (18 to 55 years) in the following ten countries: the United States, Germany, Poland, Turkey, China, India, Mexico, South Africa, Singapore, and Brazil. Additional information about our sample is provided in SI section 6.

We collected a minimum of 800 (maximum 850) respondents per country (+5–10% quota sampling allowance). A survey research company (Bilendi) was contracted to field the survey online to panelists in ten countries. All surveys were programmed in-house and hosted on Qualtrics in seven languages representative of the country of origin: English, German, Polish, Spanish, Portuguese, Turkish, and Standard Chinese. We conducted a sensitivity power analysis using the DeclareDesign simulation framework (Blair et al., 2023) and found that we are sufficiently powered both in the pooled model as well as the comparative models (Schönbrodt and Perugini, 2013). The analyses and a brief discussion can be found in SI section 8, Tables 10 and 11.

Data collection for all countries was conducted from July to November 2023. The survey was administered to 13,407 participants, some of whom were automatically excluded for non-consent, duplication, geographic restrictions, or falling outside of quota conditions. During data cleaning, additional exclusion criteria were applied to remove speeders (completion time of less than 3 min for the approximately 13 min survey) and poor-quality responses (failing commitment and pre-treatment attention checks). We also excluded incomplete survey responses where fewer than four vignettes were rated or >90% of the total responses were missing. The final sample consists of 8441 respondents across ten countries (minimum: 790 in India, maximum: 875 in Turkey, average: 844 per country).

Due to time and budgetary restraints, we relied on quota sampling with a focus on gender and age to recruit respondents through online panels. While we accept that this method may not produce highly representative samples of the population, we still expect that the criteria for our quotas limit major issues with representativeness. The quotas consisted of equal shares of males to females per four age groups from 18 to 54 years. For example, in Germany, we recruited 100 (up to a maximum of 110) female participants in the 18–24 age group. Across countries, our samples consist of respondents from different (self-assigned) social classes who are likely more active on digital channels due to the online nature of the survey. Demographics for every country, including age cohort, gender, education, employment, political orientation, social status, and religiosity, can be found in SI Table 8.

Vignette experiment

We presented respondents with a sequence of four vignettes, each from one discipline. These fictitious examples mirror short newspaper articles highlighting the scientists and their contributions. Five experimental dimensions were explored in the vignettes: Discipline (material science, genetics, economics, education research), gender (male, female), nationality (co-national, non-national), activism of the scientist (yes, no), and public impact of the research (high, low).

For example, the vignette text for material science was constructed as follows: “The [NATIONALITY] materials scientist [NAME] discovers a new material that can capture heat generated by electricity and store it in special bricks for later use. [PRONOUNS] has worked on this research for several years. [PRONOUNS] has been outspoken about the issue in public and often discusses it on social media [ACTIVISM]. This discovery could generate renewable energy for around [IMPACT] percent of residential houses in the country.” Vignette texts for all other disciplines can be found in SI section 7. Table 1 shows an overview of the vignette dimensions and levels.

Discipline

Many studies have found that scientific disciplines are perceived differently (Altenmüller et al., 2024). We chose four disciplines to represent different scientific fields: genetics, material science, economics, and education studies.

Gender

We varied the gender of the fictitious scientist using either female or male full names and pronouns which are typical for the respective country.

Nationality

We varied the nationality of the scientists by using common full names from each country the survey was fielded in. Respondents were shown either a co-national scientist or a non-national U.S. scientist. For consistency, respondents in the U.S. were shown a non-national Chinese scientist.

Activism

We introduced an experimental variation stating that the scientist in question was engaging in activism related to the scientific work.

Impact

Scientific work may be perceived differently, depending on the potential impact it has on society. We varied the impact of the scientific contribution made by the scientist by stating that it either impacted (i) very few people (between 0.01 to 0.05% affected) or (ii) many people (between 20 to 25% affected).

Since all respondents viewed one vignette for each discipline, this dimension was fixed. The other four dimensions, gender, nationality, activism of the scientist, and public impact of the research, were randomly generated for each vignette. Each of the four vignette sets consisted of 16 individual vignettes. There were 64 vignettes in total.

Trust outcomes

Respondents were asked to rate the trustworthiness of both the scientist and the science presented in each vignette. The complexity of trust has been widely recognized, with research highlighting that individuals may place greater trust in scientific principles while simultaneously having low trust in scientific institutions (Achterberg et al., 2017; Contessa, 2022; Hendriks et al., 2016; Resnik, 2011). Traditional measures of trust in science often rely on single-item surveys asking participants about their overall trust in science and research (Schoor & Schütz, 2021; Wellcome Trust & Gallup World Poll, 2018). However, these direct measures have faced criticism for not adequately capturing the nuanced value judgments that underlie trust (Besley et al., 2021).

To address this limitation, researchers have developed multiple-item scales to more effectively assess trust in science and scientists (Nadelson et al., 2014). Besley et al. (2021) conducted a comprehensive review, highlighting the proliferation of trust instruments, identifying 88 distinct measures focused on trust, credibility, and fairness in the scientific community.

Since our vignettes manipulate specific characteristics of a specific scientist in a specific field, we adopted a multi-dimensional instrument for measuring trust in scientists. In particular, we relied on the Munster Epistemic Trustworthiness Inventory (METI) from Hendriks et al. (2015). This inventory evaluates epistemic trustworthiness and source credibility across three dimensions: expertise, integrity, and benevolence. According to Hendriks et al. (2015, p. 17), a layperson’s decision to trust a specific expert (and their information) in a specific situation is based on assigning trustworthiness to these three different dimensions. While METI comprises 14 items within the three sub-scales, in the interests of parsimony, we selected three highly relevant items based on their strong loading on each dimension: “competent” (expertise), “honest” (integrity), and “responsible” (benevolence).

To measure trust in the science itself, we focused on the perceived trustworthiness of the research being conducted (Wintterlin et al., 2022). Another important dimension of trust is the individual’s perceived value of and support for scientific research in the public domain (Yin et al., 2022). To capture a comprehensive understanding of trust in science, we also measured individual support for policy influence and public funding of research.

For each vignette, respondents rated six items on a 1–4 scale (1: definitely not; 2: probably not; 3: probably yes; 4: definitely yes), as follows:

Trust in scientist

(1) Do you think the scientist is competent? (2) Do you think the scientist is honest? (3) Do you think the scientist is responsible?

Trust in science presented

(1) Do you think this research is trustworthy? (2) Do you think this research should be funded with public money? (3) Do you think this research should influence political decisions?

Relevant covariates

The pre-treatment questionnaire consisted of two sections, the first asking for demographic information and the second focusing on science education, understanding, and exposure to science. The demographics include age, gender, education level, employment, social class, religiosity, and political orientation. Here, we provide further details about the three main covariates of interest: education, political orientation, and science consumption. In our covariate analysis, we investigate how the relationship between these factors affecting trust in science varies across different levels of democracy within our country sample.

Education

We asked respondents, “What is the highest level of education you have obtained?” This categorical item was measured on a 4-point scale: 1 (less than high school), 2 (high school or further education), 3 (college or university graduate), and 4 (doctorate or equivalent).

Political orientation

Following the prevailing approach in cross-national surveys, political orientation was measured on an 11-point scale, from 1 (furthest left) to 11 (furthest right): “In political matters, people talk of the “left” and the “right”. How would you place your views on this scale, generally speaking?” In China, this scale was reversed and later recoded during preprocessing to facilitate comparison (see World Values Survey (Haerpfer et al., 2022; Beattie et al., 2022).

Science consumption

This categorical measure is an aggregate of 5 items: “How often do you do any of the following: watch documentaries about scientific topics, read non-fiction books about scientific topics, read (newspaper) articles about scientific topics, listen to radio shows or podcasts about scientific topics, and consume scientific formats on social media?” Each item was measured on a 4-point scale: 1 (almost never), 2 (about 1–2 times per month), 3 (about 1–2 times per week), and 4 (almost daily).

General trust in science

We also included pre-treatment items on general trust in science, independent of the vignettes. We use this measure as an outcome variable to examine the interaction between the main covariates described above and the regime type (see Fig. 3). We asked respondents the following: “How much do you trust … science?” This outcome was measured on a 5-point Likert scale ranging from 1 (not at all) to 5 (a great deal).

Regime type and liberal democracy index

Our aim was to examine how previously identified factors affecting public trust in science vary across different political regimes, contingent on the level of democracy. We used the Liberal Democracy Index (LDI) from V-Dem (Coppedge et al., 2023) to measure the degree of democracy within countries (see SI section 6 for democracy scores for each country in 2023). The V-Dem indices are expert-validated measures devised for each country per year and are widely trusted in political science research. Since the LDI is scaled from 0 (least democratic) to 1 (most democratic), following general practice, we use 0.5 as the cut-off for non-democratic countries.

Statistical methods

To analyze our vignette survey experiment, we used mixed-effect models with respondent and country random intercepts for pooled analyses, and only respondent random intercepts for independent country-level analyses (Baguley et al., 2021). For all vignette outcomes, we employed linear model specifications. Hypotheses were tested by regressing the trust outcomes on the treatment indicators (Hainmueller et al., 2014), without including any other covariates. Mixed-effect models were fit using the lme4 R package. Since we employed effect-coding and suppressed the intercept, the estimate for each discipline represents the mean value of the trust outcome for that discipline.

For the covariate analyses, we used pre-treatment measures of education, consumption of science media, and political orientation. The outcome variable is the pre-treatment level of trust in science. For these linear models, we interacted the pre-treatment covariates with the liberal democracy index from the V-Dem dataset (Coppedge et al., 2023) to approximate the level of democracy within political regimes. Age and employment were included as controls in all three interaction models.

Pooled effects of scientist and science Characteristics

Considering the results of the vignette experiment across the complete sample, we find that people’s trustworthiness assessments vary considerably depending on the specific characteristics of the scientist and the scientific findings presented (see results in Fig. 1). Regarding characteristics of the scientist, against expectations, people rate male scientists lower than female scientists across all person dimensions: as less competent (β = −0.020; p < 0.01) and less responsible (β = −0.024; p < 0.01), and slightly significantly less honest (β = −0.014; p < 0.05). This effect extends to male scientists’ research, in which participants prefer to be less supported with public funding (β = −0.030; p < 0.01), to have less influence on policy (β = −0.023; p < 0.01), and they perceive as less trustworthy overall (β = −0.020; p < 0.01), compared to research conducted by female scientists.

Contrary to our expectations, scientists presented as engaging in activism are perceived as more honest compared to non-activist scientists (β = 0.028; p < 0.01), and participants are more in favor of their research influencing policy (β = 0.020; p < 0.05). However, there are no other significant effects of activism. Co-nationality only increases the perceived responsibility of the scientist (β = 0.015; p < 0.05), with no significant effects on science trustworthiness outcome measures. Additionally, whether a scientist is a co-national or a U.S. citizen does not influence respondents’ assessments of honesty and competence. For the U.S. sample, the foreign scientist was framed as a Chinese citizen.

Regarding the public impact of science, when findings were presented as impacting a larger share of the population (20 − 25%) rather than a small share (0.1–0.5%), participants rated the science as more trustworthy (β = 0.039), the scientist as more competent (β = 0.043), responsible (β = 0.046), and honest (β = 0.027), and are significantly more in favor of the science being supported by public funding (β = 0.145) and having policy impact (β = 0.125). While these effects vary in size, they are all highly positively significant (p < 0.001). This indicates that respondents do not seem to separate their perceptions of scientists and science from the potential public impact this research may have, providing higher ratings when this impact is greater.

Finally, there are considerable differences in the perception of scientific disciplines. As expected, participants rate genetics as the most trustworthy, followed by material science (a sub-discipline of engineering), then economics and education (β_genetics = 3.00, β_material = 2.93, β_economics = 2.78, β_education = 2.78, respectively; p < 0.01 for all). These relatively large effect-coded coefficients indicate the mean trust outcome for each discipline, allowing for direct comparisons between them. For example, β_genetics is 3.06 (p < 0.01) for the honest outcome measure, meaning that geneticists are perceived as significantly more honest than the average perception of honesty across all other disciplines.

However, even though people show high trust in scientists and science in the case of genetics and material science, they believe this research should have less policy influence than economics. While education research scores lowest among our selection of scientific disciplines, scientists in this field are rated as equally honest as economists, and education research is perceived as about as trustworthy as economics. The results of the discipline treatment appear intuitive with respect to the public stereotypes of the perceived “scientificness” of different disciplines. This ranking is mirrored by the participants’ ranking of a greater set of disciplines in the survey module of the study (see more details in SI section 6). While the discipline effect within the vignette study serves as a sanity check, it cannot be disentangled from the fictitious, discipline-specific research descriptions presented in the vignettes (see SI section 7 for the full text).

Cross-national comparisons of scientist and science characteristics

Do the experimental effects on scientists and science ratings differ depending on regime type? To report vignette experiment results separately for different countries, we aggregate the six outcome measures to two subordinate categories: trust in scientist as the average of competence, honesty, and responsibility; and trust in science as the average of science trustworthiness, support for public funding of science, and support for science’s impact on policy (see Fig. 2).

Across the countries in our sample, we find largely consistent impacts of science and scientist characteristics on public trust in science and scientists across different countries and political regimes. The effect of gender appears to be primarily driven by a few countries that cover the entire range of the democracy index: Germany and Mexico, where there is lower trust in male scientists and the science they produce (β = −0.056, −0.059, respectively; p < 0.01), and China, where male scientists are rated less favorably (β = −0.041; p < 0.05). The effect of public activism is driven by Singapore, where activist engagement leads to greater trust in the scientist (β = 0.047; p < 0.01), and Turkey, where trust in science is higher for activists (β = 0.059; p < 0.01). In contrast, research by scientists engaging in public activism is rated as significantly less trustworthy in Poland (β = −0.042; p < 0.05). In all other countries, we do not find statistically significant effects of gender or activism. The null effect of scientists’ nationality is robust across all countries. For brevity, we present coefficient plots in Fig. 2, with further details on statistical inference in SI section 3, Tables 2 and 3.

The only instance in which we see systematic heterogeneity between political regimes is the effect of scientific impact. Across all countries, high-impact research is trusted more compared to low-impact research. This support for high-impact science, however, is highest in full democracies such as Germany and the United States (β = 0.195, 0.172; p < 0.01) and weaker in autocracies such as China and Turkey (β = 0.095, 0.056; p < 0.01). While the size of discipline effects differs substantially between countries, the direction, and order is consistent with few exceptions. Education researchers and their research are rated least favorably across countries, except in Germany, where economists are rated even lower (β = 2.75; p < 0.01), although this disdain does not extend to economics research (β = 2.51; p < 0.01). Overwhelmingly, geneticists and material scientists are rated most favorably across countries. Regarding their research, we observe some heterogeneity, which, however, does not systematically co-vary with the political regime. Overall, public perceptions of scientists and science appear to be largely consistent across countries with different political regimes.

The role of education, science consumption, and political orientation

To explore effect heterogeneity, we now consider the respondent characteristics of education, consumption of science media, and political orientation. First, we ask: How does the relationship between education and trust in science play out in different political regimes? We find that political regime type does not influence the relationship between education and trust in science (see Fig. 3, Panel A). While, in general, people with higher education show greater trust in science, this difference is less pronounced for people in liberal democracies (predicted trust: 3.95 in non-democracies, 3.84 in democracies). This mirrors findings from observational survey data on a global sample of countries (Jiang & Wan, 2023), and we find no support for the positive effect of democracy on the trust–education relationship. In fact, highly educated individuals in non-democratic countries have significantly higher levels of trust compared to all other groups. Estimating the relationship between education and trust in science separately confirms the consistency of this relationship across political regimes (SI section 4).

Next, we examine how the relationship between political orientation and trust in science differs across political regimes and countries. Figure 3 (Panel B) indicates that predicted trust is much higher in non-democracies for both right and left-leaning respondents than in democracies. The gap is more pronounced for right-leaning individuals who have significantly higher trust in science than their counterparts in democracies (predicted trust: 3.81 in non-democracies, 3.59 in democracies) while left-leaning individuals have a much smaller difference (3.85 in non-democracies, 3.76 in democracies). While we find that the difference in trust between left- and right-leaning individuals is strongly significant within democracies, there is no significant difference in trust between the two camps within non-democracies. In other words, our results show that political orientation has a much stronger effect on public perceptions in democratic countries than in autocratic ones. In addition, we find a significant interaction between political orientation and democracy, such that an increase in democracy level (continuously measured from 0 to 1) reduces trust in science (see SI section 5 Table 5 for interaction results). Investigating this relationship at the level of individual countries, we find negative associations between right-leaning political orientation and trust in science in most countries, including the edges of the democracy index (see SI section 4). The opposite is true only for India, where left-leaning political orientation is associated with science skepticism. In Singapore, Poland, Mexico, and South Africa, the relationships are less pronounced.

Finally, we consider how the relationship between science consumption and trust in science differs across political regimes. We focus on primary exposure resulting from active consumption of science media based on respondents’ self-reported habits of engagement with science news, podcasts, documentaries, social media, etc. Although the difference in trust between high consumers (3.96 in non-democracies, 3.88 in democracies) and low consumers (3.62 in non-democracies, 3.52 in democracies) is significant within non-democracies and democracies alike, political regime type does not moderate this relationship. However, high and low consumers in non-democracies have higher levels of trust than their democratic counterparts. When we break down these results for individual countries (SI section 4), we find that non-democracies such as China, India, and Turkey display a linear relationship between increased consumption and higher trust. In contrast, for the U.S., Germany, and Brazil, the relationship takes a quadratic form, as trust levels are highest for moderate consumers of science media. The relationship is much less clear in hybrid regimes like Singapore and Mexico.

In this cross-national survey experiment, we systematically evaluated public perceptions of scientists as persons and the science they produce. More specifically, we shed light on the causal effects of different science and scientist characteristics on different dimensions of trust. To achieve this, we constructed vignettes across a wide range of scientific disciplines, actively avoiding publicly salient and politicized scientific issues (e.g., climate change, COVID-19). Across the entire spectrum of political regimes, we find largely consistent effects of scientist and science characteristics on public trust. Effects of the scientist’s gender and activism or the scientific impact and the discipline of the presented research appear parallel between democracies and non-democracies. Examining respondent characteristics more closely, we find considerable homogeneity in trust in science for education and science consumption. In stark contrast, the relationship between political orientation and trust in science is heavily moderated by political regimes: while right-leaning individuals in democracies show significantly lower trust in science than left-leaning individuals, there is no such difference in non-democracies. Finally, regarding the six vignette response dimensions of trust in science, as well as the pre-treatment survey measure, respondents in non-democratic countries display greater trust in science than those in democracies.

Beyond our core research question, our findings reveal several interesting insights. Contrary to expectations, female scientists are rated more competent, honest, and responsible than male scientists—a finding that is not specifically driven by countries with more emancipated societies. In line with other recent evidence (Altenmüller et al., 2024), the common expectation that men are trusted more than women no longer applies, at least in the traditionally male-dominated domain of science. In other domains of trust in experts, several parallel effects were found. For example, people are more likely to trust female physicians compared to their male counterparts, especially if they are female themselves (Karafillakis et al., 2022). The mechanism of this effect remains to be explored in future research. For example, the notion that assumed “caring” leadership styles in female government officials lead to greater trust in female leaders did not hold as explanation in related work (Willis et al., 2021).

Furthermore, the positive or neutral (but certainly not negative) effects of scientists’ public activism on their trustworthiness encourage high-quality science communication and scientists’ involvement in informing political decision-making. In line with our findings, previous research found no decline in climate researchers’ credibility when they engaged in advocacy (Kotcher et al., 2017). Nevertheless, we emphasize that politicization of scientific issues can still occur, and it may have implications for the scientists involved in highly politicized debates (Cofnas et al., 2018). In our study, we explicitly selected low-salience scientific issues for our vignettes, also since more politicized domains have been examined in related research (Gauchat, 2012; Schrøder, 2022). In general, however, across the non-democratic countries in our sample, science appears to be less politicized than in democracies since we find little to no political polarization regarding general levels of trust in science in these countries.

Both the characteristics of scientists as persons but also of the science they are involved in have consistent effects on public perceptions and trust. While it seems intuitive that high-impact research is trusted more than low-impact research, this also implies that people want science to have a positive impact on their lives. It does not seem that the public is predominantly skeptical or even scared of the potential risks of high-impact research. Across political regimes, however, the magnitude of this effect differs, and we find that respondents in democracies seem to base their trust in science much more on the political impact that this science may have. We find similar evaluations of different scientific disciplines, with “hard” sciences being perceived as more trustworthy, while other disciplines, in particular the social sciences, receive lower ratings. However, a limitation of our design is that the disciplines cannot be separated from the findings presented in the vignettes. Future research may overcome this by varying findings within discipline or even discipline within findings.

While the statistical effects identified in our study are modest, they should not be dismissed as inconsequential. Small effects can accumulate to produce significant real-world impacts. When considering entire populations, these effects translate into millions of individuals with similar perceptions, attitudes, and behaviors toward scientists and scientific research. This context highlights the importance of our findings: for instance, lower trust and perceived value of ‘soft’ science disciplines can significantly affect support for public funding and policy impact. Conversely, greater trust in high-impact research can improve the success of public health campaigns, environmental regulations, and educational programs that are more effectively communicated to the public.

Our study also finds consistent effects across countries, with differences in outcomes providing valuable insights for developing targeted interventions. Policymakers and science communicators can leverage these effects to enhance public engagement and compliance with evidence-based policies. It is crucial to consider the broader context in which these effects operate. Societal factors, such as political climate and educational systems, may amplify or mitigate the observed effects. By identifying specific demographics or country contexts where certain effects are more pronounced, stakeholders can design more effective strategies to challenge preconceived notions, reduce the influence of political biases, and improve public trust in science.

With our cross-national design, we set out to determine how science and scientists are perceived in countries over the entire spectrum of political regimes. Overall, we find great consistency in these perceptions and trust ratings. This is good news. Science is a collaborative endeavor that requires scientific cooperation across national borders, and we show that public perceptions of this endeavor are similar, even across countries with very different political environments. Nevertheless, while this finding is encouraging, we should not be too optimistic when it comes to the role of science in fostering societal progress. One of the main barriers is the increasing politicization of science. In autocracies, scientific agendas are still strongly shaped by political regimes. Governments determine which research is funded and who contributes to it. This means that in many countries worldwide, science is not free, and many issues will fail to be addressed because they are politically contentious. In democracies, by contrast, the problem lies with the politicized public perception of science. Our results show that political orientation strongly determines people’s assessment of scientists and science. Clearly, science no longer holds a universal perception as an independent and neutral arbiter of truth, which threatens its important role in society.