Content uploaded by Ani Bajrami
Author content
All content in this area was uploaded by Ani Bajrami on Feb 05, 2024
Content may be subject to copyright.
This is a preprint and is undergoing peer review.
1
Trust in scientists and their role in society across 67 countries
Authors: Viktoria Cologna1*, Niels G. Mede2, Sebastian Berger3, John Besley4, Cameron
Brick5,6, Marina Joubert7, Edward W. Maibach8, Sabina Mihelj9, Naomi Oreskes1, Mike S.
Schäfer2, Sander van der Linden10, Nor Izzatina Abdul Aziz11, Suleiman Abdulsalam12,
Nurulaini Abu Shamsi13, Balazs Aczel14, Indro Adinugroho15,16, Eleonora Alabrese17, Alaa
Aldoh5, Mark Alfano18, Mohammed Alsobay19, Marlene Altenmüller20, R. Michael Alvarez21,
Richard Amoako22, Tabitha Amollo23, Patrick Ansah22, Denisa Apriliawati24, Flavio Azevedo25,
Ani Bajrami26, Ronita Bardhan27, Keagile Bati28, Eri Bertsou29, Cornelia Betsch30, Apurav Yash
Bhatiya31, Rahul Bhui19,32, Olga Białobrzeska33, Michał Bilewicz34, Ayoub Bouguettaya35,
Katherine Breeden36, Amélie Bret37, Ondrej Buchel38, Pablo Cabrera-Álvarez39, Federica
Cagnoli40, André Calero Valdez41, Timothy Callaghan42, Rizza Kaye Cases43, Sami Çoksan44,
Gabriela Czarnek45, Steven De Peuter46, Ramit Debnath21,47, Sylvain Delouvée48, Celia Díaz-
Catalán39, Lucia Di Stefano40, Kimberly C. Doell49, Simone Dohle50, Karen M. Douglas51,
Charlotte Dries52, Dmitrii Dubrov53, Małgorzata Dzimińska54, Ullrich K. H. Ecker55, Christian T.
Elbaek56, Mahmoud Elsherif35, Benjamin Enke57, Tom W. Etienne58, Matthew Facciani59,
Antoinette Fage-Butler60, Md. Zaki Faisal61, Xiaoli Fan62, Christina Farhart63, Christoph
Feldhaus64, Marinus Ferreira18, Stefan Feuerriegel65, Helen Fischer66,67, Jana Freundt68, Malte
Friese69, Simon Fuglsang70, Albina Gallyamova53, Patricia Garrido-Vásquez71, Mauricio Garrido
Vásquez71, Winfred Gatua72, Oliver Genschow73, Omid Ghasemi74, Theofilos Gkinopoulos45,
Jamie Gloor75, Ellen Goddard62, Mario Gollwitzer20, Claudia González Brambila76, Hazel
Gordon15, Dmitry Grigoryev53, Gina M. Grimshaw77, Lars Guenther78, Håvard Haarstad79, Dana
Harari80, Lelia N. Hawkins81, Przemysław Hensel82, Alma Cristal Hernández-Mondragón83, Atar
Herziger80, Guanxiong Huang84, Markus Huff66, Mairéad Hurley85, Nygmet Ibadildin86, Maho
Ishibashi87, Mohammad Tarikul Islam88, Younes Jeddi12, Tao Jin89, Charlotte A. Jones90,
Sebastian Jungkunz91, Dominika Jurgiel92, Zhangir Kabdulkair86, Jo-Ju Kao93, Sarah
Kavassalis81, John R. Kerr94, Mariana Kitsa95, Tereza Klabíková Rábová96, Olivier Klein97,
Hoyoun Koh98, Aki Koivula99, Lilian Kojan41, Elizaveta Komyaginskaya53, Laura König100, Lina
Koppel101, Alexandra Kosachenko102, John Kotcher8, Laura S. Kranz77, Pradeep Krishnan29, Silje
Kristiansen103, André Krouwel104, Toon Kuppens105, Eleni A. Kyza106, Claus Lamm49, Anthony
Lantian107, Aleksandra Lazić108, Oscar Lecuona de la Cruz109, Jean-Baptiste Légal107, Zoe
Leviston110, Neil Levy18, 111, Amanda Lindkvist101, Grégoire Lits112, Andreas Löschel64, Alberto
López Ortega104, Carlos Lopez-Villavicencio113, Nigel Mantou Lou114, Chloe H. Lucas90, Kristin
Lunz-Trujillo115,116, Mathew D. Marques117, Sabrina Mayer91, Ryan McKay118, Hugo Mercier119,
Julia Metag120, Taciano L. Milfont121, Joanne Miller122, Panagiotis Mitkidis56, Fredy Monge-
Rodríguez113, Matt Motta42, Iryna Mudra95, Zarja Muršič123, Jennifer Namutebi124, Eryn J.
Newman110, Jonas P. Nitschke49, Fernando Luiz Nobre Cavalcante125, Daniel Nwogwugwu126,
Thomas Ostermann127, Tobias Otterbring128, Jaime Palmer-Hague129, Myrto Pantazi97, Philip
Pärnamets130, Paolo Parra Saiani40, Mariola Paruzel-Czachura131,132, Michal Parzuchowski33,
Yuri Pavlov102, Adam R. Pearson133, Myron A. Penner129, Charlotte R. Pennington134, Katerina
This is a preprint and is undergoing peer review.
2
Petkanopoulou135, Marija Petrović108, Jan Pfänder119, Dinara Pisareva98, Adam Ploszaj136,
Karolína Poliaková96, Ekaterina Pronizius49, Katarzyna Pypno131, Diwa Malaya Quiñones137,
Pekka Räsänen99, Adrian Rauchfleisch93, Felix G. Rebitschek52,138, Cintia Refojo Seronero39,
Gabriel Rêgo139, James P. Reynolds134, Joseph Roche85, Simone Rödder140, Jan Philipp Röer127,
Robert M. Ross18, Isabelle Ruin141, Osvaldo Santos142, Ricardo R. Santos142, Philipp Schmid30,
143, 144, Stefan Schulreich145,146, Bermond Scoggins147, Amena Sharaf148, Justin Sheria
Nfundiko149, 150, Emily Shuckburgh47, Johan Six151, Nevin Solak148, Leonhard Späth151, Bram
Spruyt152, Olivier Standaert112, Samantha K. Stanley110, Gert Storms46, Noel Strahm3, Stylianos
Syropoulos153, Barnabas Szaszi14, Ewa Szumowska45, Mikihito Tanaka154, Claudia Teran-
Escobar141, Boryana Todorova49, Abdoul Kafid Toko12, Renata Tokrri155, Daniel Toribio-
Florez51, Manos Tsakiris118, 156, Michael Tyrala157, Özden Melis Uluğ158, Ijeoma Chinwe
Uzoma159, Jochem van Noord25, Christiana Varda106,160, Steven Verheyen161, Iris Vilares89,
Madalina Vlasceanu162, Andreas von Bubnoff163, Iain Walker110, 164, Izabela Warwas54, Marcel
Weber69, Tim Weninger59, Mareike Westfal73, Florian Wintterlin120, Adrian Dominik Wojcik165,
Ziqian Xia166, Jinliang Xie167, Ewa Zegler-Poleska140, Amber Zenklusen29, Rolf A. Zwaan161
*Corresponding author. Email: v.cologna@ikmz.uzh.ch
This is a preprint and is undergoing peer review.
3
Abstract
Scientific information is crucial for evidence-based decision-making. Public trust in science can
help decision-makers act based on the best available evidence, especially during crises such as
climate change or the COVID-19 pandemic1,2. However, in recent years the epistemic authority
of science has been challenged, causing concerns about low public trust in scientists3. Here we
interrogated these concerns with a pre-registered 67-country survey of 71,417 respondents on all
inhabited continents and find that in most countries, a majority of the public trust scientists and
think that scientists should be more engaged in policymaking. We further show that there is a
discrepancy between the public’s perceived and desired priorities of scientific research.
Moreover, we find variations between and within countries, which we explain with individual-
and country-level variables, including political orientation. While these results do not show
widespread lack of trust in scientists, we cannot discount the concern that lack of trust in
scientists by even a small minority may affect considerations of scientific evidence in
policymaking. These findings have implications for scientists and policymakers seeking to
maintain and increase trust in scientists.
Introduction
Public trust in science provides many benefits to people and society at large. It helps people
make informed decisions (e.g., on health and nutrition) based on the best available evidence,
provides the foundation for evidence-based policymaking, and warrants government spending on
research. Trust in science is also vital for the management of global crises like the COVID-19
pandemic and climate change. Societies with high public trust in science dealt with the COVID-
19 pandemic more effectively, as citizens were more likely to comply with non-pharmaceutical
COVID-19 interventions2 and had higher vaccine confidence1. People with high trust in science
are also more likely to engage in individual and collective action on climate change4,5.
Globally, most people trust science, and scientists are among the most trusted actors in
society6–8. However, previous studies point to strong national and regional differences in trust in
science, with anti-science attitudes and lower trust in science being more prevalent in some Latin
American and African countries7,9,10. To avoid misleading inferences from Western to non-
Western countries, more global, comparative research on trust in science is imperative.
While science is generally held in high esteem, its epistemic and cultural authority has
been challenged by mis- and disinformation11,12, historical failings of science13, an alleged
This is a preprint and is undergoing peer review.
4
“reproducibility crisis”14, conspiracy theories15,16, and science-related populist attitudes17,18.
Science-related populism has been conceptualised as a perceived antagonism between ‘the
ordinary people’ and common sense versus academic elites and scientific expertise17. Unlike
political populism, which criticises political elites and their political power claims, science-
related populism criticises academic elites, challenges their decision-making authority in
scientific research, and suggests that their epistemic truth claims are inferior to the common
sense of ‘the people’17. Anti-science attitudes, even if held by only a minority of people, raise
concerns about a potential crisis of trust in science which could challenge the epistemic authority
of science and the role of scientists in supporting evidence-based policymaking9,17. These
concerns, which have been prominently discussed in leading news media, have been exacerbated
as trust in scientists and their desired role in policymaking has become divided along partisan
lines. A number of previous studies show that in the US and some other countries, conservatives
and right-leaning individuals have low levels of trust in scientists, hold stronger anti-science
attitudes, and express low confidence that scientists act in the best interest of the public, provide
benefits to society, and apply reliable methods9,10,19–21.
Overall, however, there is scant robust global comparative evidence on trust in science,
and, in particular, the extent to which concern over a lack of trust in scientists is justified. This
raises the risk of ill-informed science policies, and misconceptions about the state of science in
society. These concerns call for a global study on the prevalence and correlates of trust in
scientists and public expectations about scientists’ role in society and policymaking.
There are only a few global studies on trust in scientists, and they have significant
limitations6–10,22,23. They either focus on Western countries, assess a limited range of theoretical
constructs, or do not assess normative perceptions of the role of scientists in society and
policymaking—a central construct in our study. Our large-scale, pre-registered survey addresses
these limitations by offering the first global dataset on trust in scientists post-Covid-19
pandemic24; being the first global study to investigate normative perceptions of scientists in
policymaking; using a theoretically informed multidimensional trust measure25; examining
demographic, attitudinal, and country-level factors that explain why trust varies across countries;
surveying underrepresented countries and individuals in research26; and, in almost all countries,
including local collaborators27,28. Our survey also goes beyond commonly studied correlates of
trust in scientists by investigating people’s desired research priorities.
This is a preprint and is undergoing peer review.
5
Our study was a crowd-sourced Many Labs project with the same translated online
questionnaire given to 71,417 respondents in k = 67 countriesi on all inhabited continents (fig.
S1
1
). The survey covered 31% of the world’s countries that jointly make up 78% of the global
population. Data were collected between November 2022 and August 2023, with quota samples
that were weighted according to national distributions of age, gender, and education level, as
well as country sample size. As recommended by other global studies on trust in science7, we
provided respondents with a definition of science and scientists to mitigate semantic variations
across languages (see SI). For example, “science” does not translate precisely into German,
Swedish, and Polish, where the term also encompasses the “humanities”. We slightly deviated
from the preregistration because of multicollinearity (exclusion of confidence in science as a
model covariate) and sparsely populated sample strata in certain countries (collapsing
neighbouring strata during post-hoc weighting; see SI).
Trust in scientists across the world
We assessed trust in scientists with an index composed from a 12-item scale measuring four
established dimensions of trustworthiness: perceived competence, benevolence, integrity, and
openness25,29,30. This is based on the most comprehensive review of trust measures used to assess
perceptions of scientists30; was pre-tested to confirm its reliability; relies on accepted conceptual
assumptions that we validated in factor analyses; and has high reliability across countries (see
SI). However, confirmatory factor analyses show that we can only assume configural invariance
and no metric or scalar invariance (see SI). This is a common caveat of multilingual survey
research and to some extent unavoidable31.
Globally, trust in scientists is moderately high (global M = 3.62, SD = 0.70; 1 = very low,
3 = neither high nor low, 5 = very high). No country surveyed revealed low trust in scientists
overall (Fig. 1). Across the globe, people perceive scientists as having high competence (M =
4.02, SD = 0.71), moderate integrity (M = 3.59, SD = 0.78), and benevolent intentions (M = 3.55,
SD = 0.82; table S1). Scientists’ perceived openness to feedback is slightly lower (M = 3.33, SD
= 0.86), with 23% believing that scientists pay only somewhat or very little attention to others’
views (fig. S2). Globally, 75% agree that scientific research methods are the best way to find out
if something is true or false. Trust in scientific methods moderately correlates with trust in
1
Supplementary Information is available from the corresponding author upon request.
This is a preprint and is undergoing peer review.
6
scientists (r = 0.48, p < .001), supporting previous findings on the multidimensionality of trust in
science32.
Fig. 1. Weighted means (M) for trust in scientists across countries and regions (1 = very
low, 3 = neither high nor low, 5 = very high). Note. Vertical line denotes weighted global
mean. Horizontal lines indicate standard errors (SE). Country-level SEs range between 0.008-
0.133.
This is a preprint and is undergoing peer review.
7
Country-level analyses show that trust in scientists differs considerably across countries
and world regions (Fig. 1). For example, trust is highest in Egypt and India and lowest in Albania
and Kazakhstan. Contrary to previous studies7,8, we do not find a clear pattern that scientists are
less trusted in Latin American and African countries. However, we do find patterns within
specific regions. For example, Russia as well as several former Soviet republics and satellite
states show relatively low trust in scientists.
Explanatory factors of trust in scientists and cross-country variations
To identify correlates of trust in scientists globally, we fitted linear random-intercept regression
models that included post-stratification weights to provide estimates that were nationally
representative in terms of gender, age, and education in almost all countries. Women, older
people, residents of urban (versus rural) regions, people with higher income, as well as more
religious, educated, liberal, and left-leaning people trust scientists more (Fig. 2, see also table
S2). Differences across countries and sociodemographic groups can be explored with an online
dashboard developed for this project.
We find a small positive relationship between tertiary education and trust in scientists
across countries (b = 0.029, p < .001). Tertiary education has the strongest association with trust
in scientists in the United States and Austria (fig. S3). In most other countries, we find no
significant relationship between tertiary education and trust.
Globally, religiosity is positively associated with trust in scientists (b = 0.048, p < .001),
but there are substantive differences depending on which particular religion is involved. In
Muslim countries such as Türkiye, Bangladesh, and Malaysia (fig. S4), people did not perceive
strong conflicts between scientific and religious epistemologies, and trust is positively associated
with religiosity33. In contrast, religiosity is negatively related to trust in scientists in Israel and
many Christian-majority European countries, in which a majority perceives disagreements
between science and the teachings of their religion33.
This is a preprint and is undergoing peer review.
8
Fig. 2. Standardised estimates of weighted blockwise multilevel regression model testing the
association of trust in scientists with demographic characteristics, attitudes, and country-
level indicators (random intercepts across countries). Block 1 uses data from all 67 countries,
block 2 uses data from 66 countries (all except Mexico), block 3 uses data from 52 countries (all
except those where PISA’s literacy scores were not available, see supplementary material). Full
regression results are reported in table S2. ** p < .01, *** p < .001.
This is a preprint and is undergoing peer review.
9
Previous studies, mostly focused on North America and Europe, have found right-leaning
political orientation to be negatively associated with trust in scientists9,21. However, we find the
relationship to be more complex. Overall, trust in scientists is slightly higher among people with
left-leaning political orientations (b = -0.010, p = .003) than right-leaning orientation. However,
this relationship varies substantially across countries (Fig. 3). Right-leaning political orientation
is negatively associated with trust in scientists in several European and North American
countries, as well as in Brazil and Israel, so previous research, which has disproportionally
focused on these countries, has tended to stress right-leaning distrust. However, in most countries
(k = 41), our data do not show a relationship between political orientation and trust in scientists.
Furthermore, in some Eastern European, South-East Asian, and African countries, right-leaning
individuals have higher trust in scientists.
These contrasting findings may be explained by the fact that in some countries right-
leaning parties may have cultivated reservations against scientists among their supporters, while
in other countries left-leaning parties may have done so34 (see fig. S5). In other words, the
attitudes of political leadership rather than peoples’ political orientation may better explain
politically correlated attitudes towards scientists (see SI for selected country-specific
explanations). In any case, extrapolating findings from Western to non-Western countries may be
misleading.
This is a preprint and is undergoing peer review.
10
Fig. 3. Relationship of left-right political orientation and trust in scientists. Figure visualises standardised random slopes for political
orientation (1 = left – 5 = right), which were extracted from a weighted linear multilevel regression model that explained trust in
scientists (1 = very low, 3 = neither high nor low, 5 = very high) across countries and contained random intercepts and slopes of political
orientation across countries. Countries with significant effects (p < .05) are displayed in colours: Countries coloured in shades of blue
show a positive association of left-leaning orientation and trust in scientists (i.e., right-leaning have lower trust). Countries coloured in
shades of red show a positive association of right-leaning orientation and trust in scientists (i.e., left-leaning have lower trust) Countries
with non-significant effects are shaded in dark grey. Countries with no available data are shaded in light grey.
This is a preprint and is undergoing peer review.
11
Trust is significantly associated with attitudes towards science. We find positive relationships
between people’s trust in scientists and their willingness to rely on scientific advice and thus
make themselves vulnerable to scientists, the belief that science benefits people like them, and
trust in scientific methods. We also find that science-related populist attitudes—that is, beliefs
that people’s common sense is superior to the expertise of scientists and scientific institutions—
are associated with lower trust in scientists. We also tested pre-registered hypotheses assuming
that trust in scientists is linked to country-level indicators, including GDP per capita, PISA’s
science literacy score, and the Academic Freedom index.
Contrary to the finding of the Wellcome Global Monitor7, we find that trust is positively
associated with the Gini index (i.e., trust is higher in countries with more income inequality).
One possible explanation for the discrepancy between the two studies is that urban populations—
which are more likely to trust scientists (Fig. 2)—were overrepresented in our samples from
countries with high Gini scores, e.g., South Africa and Argentina. However, we suggest that the
discrepancy may not just be methodological: people in countries with high inequality may see
scientists as a trustworthy alternative to perceivably corrupt political and economic elites35–37.
Mapping average trust levels against the Gini index (fig. S6) shows that the relationship between
the two seems to be driven by countries scoring relatively high on Transparency International’s
Corruption Perceptions Index38, primarily Latin American countries as well as Sub-Saharan
African countries. We do not find that trust is higher in countries with higher science literacy
scores and government expenditures on education, which challenges assumptions that public
understanding of science and policy measures to increase it foster trust in scientists39.
Normative perceptions of scientists’ role in society and policymaking
Left-right divides in public opinion about science often centre on the question of whether
scientists should take an active role in policymaking40. We find that the level of agreement with
statements that scientists should engage in society and policymaking is moderately high (global
M = 3.64, SD = 0.87, 1 = strongly disagree – 5 = strongly agree). Globally, a large majority
(82%) agrees that scientists should communicate about science with the public, particularly in
African countries. Globally, only a minority disagrees that scientists should actively advocate for
specific policies (23%), communicate their findings to politicians (20%) and be more involved in
This is a preprint and is undergoing peer review.
12
the policymaking process (21%). However, perceptions differ across countries, with lowest
agreement in Serbia and Slovenia (fig. S7).
About a quarter of the global sample selected the scale midpoints, therefore neither
agreeing nor disagreeing on whether scientists should be more involved in policymaking and
society (Fig. 4). People with high trust in scientists strongly favour scientists’ engagement in
society and policymaking (b = 0.273, p < .001), especially in English-speaking countries
including the United States and Australia (fig. S8). Support for scientists’ engagement in society
and policymaking also varies both between and within countries. We find that people who are
younger, have tertiary education and higher income, or live in urban areas, generally approve of
scientists’ engagement in policymaking (table S3). We also find that right-leaning people and
conservatives disapprove of scientists’ engagement in society and policymaking.
Fig. 4. Weighted response probabilities for single items measuring normative perceptions
of scientists in society and policymaking.
This is a preprint and is undergoing peer review.
13
Desired research priorities and perceptions that these are tackled
We hypothesised that trust in scientists relates to another normative belief about science:
Expectations about which societal goals scientists should prioritise41. We tested this and
compared whether people’s expectations match their perceptions of whether scientists actually
tackle these goals.
Globally, people assign the highest priority to improving public health (M = 4.49, SD =
0.84, 1 = low – 5 = high), followed by solving energy problems (M = 4.38, SD = 0.91), and
reducing poverty (M = 4.08, SD = 1.10). Responses suggest a substantial discrepancy between
what they think scientists should prioritise (i.e., desired priorities) and what they perceive science
is currently prioritising (i.e., perceived priorities), with poverty reduction showing the most
substantial discrepancy (Fig. 5). The least desired research goal is developing defence and
military technology (M = 3.10, SD = 1.36). Again, there are large differences between global
regions (ranging from M = 1.88, SD = 1.21 in Uruguay to M = 4.07, SD = 1.52 in the Democratic
Republic of Congo). In African and Asian countries, people often demand high priority for
developing defence and military technology, as opposed to people in most European and Latin
American countries (fig. S9). Overall, people tend to think science prioritises developing defence
and military technology more than they desire.
This is a preprint and is undergoing peer review.
14
Fig. 5. Perceived research priorities for four goals of scientific research (blue) and desired
research priorities (red). Grey horizontal lines indicate the discrepancy between desired
priorities “Scientists should prioritize this goal”) and perceived research priorities (“Science aims
to tackle this goal”). Stars indicate results of weighted paired-samples t-tests for significant
differences between perceptions and expectations, *** p < .001.
Trust in scientists is strongly associated with the discrepancy between people’s desired research
priorities and perceptions that science aims to tackle them (table S4). For people with higher trust
in scientists, odds are higher that perceived priorities exceed desired priorities. This applies also
to younger people, males, and people with right-leaning political orientation. However, people
with lower trust in scientists, left-leaning views, and liberal political orientation are more likely
to perceive that science prioritises developing defence and military technology development
more than they desire. This is in line with existing research42.
Discussion and recommendations
Our global, 67-country survey challenges concerns of a widespread lack of public trust in
scientists. In most countries surveyed, scientists and scientific methods are trusted. Globally,
This is a preprint and is undergoing peer review.
15
scientists’ perceived competence, benevolence, and integrity are high, but perceptions of
scientists’ openness are comparably lower. Therefore, scientists wishing to gain more public trust
could work on being more receptive to feedback, transparent about their funding and data
sources, and invest more effort into the types of public communication desired by a large
majority of the public. Since many people think that scientists do not pay much attention to other
people’s views, we recommend avoiding top-down communication but encouraging public
participation in genuine dialogue, in which scientists seek to consider the insights and needs of
other societal actors43. While our cross-sectional data cannot speak to the causality of these
recommendations, they are supported by the extensive literature on the determinants of trust in
scientists (see e.g., 25,29,44).
We find considerable differences of trust and its dimensions between and within
countries, which demonstrate the importance of using multidimensional trust measures in
comparative survey research25. It should be noted that our study assessed trust in scientists
without distinguishing between different fields. In some countries, trust may depend on the
scientists’ discipline and the potential impacts of science on public policy45,46.
While we find that no country has low trust in scientists on average, lack of trust in
scientists by even a small minority needs to be taken seriously, as it may affect considerations of
scientific evidence in policymaking, as well as decisions by individuals that can affect society at
large. Future research should investigate the size of these distrusting minorities across countries
and their characteristics.
Not only how much trust but also its correlates, such as right-leaning political orientation,
education, and religiosity, vary clearly across countries. This exemplifies the need for more
international research that includes underrepresented countries and understudied subpopulations.
Given these findings, we urge scholars to be cautious when generalising findings from Western
or Anglophone countries.
In nearly all countries, a majority of people want scientists to take part in policymaking.
Future global comparative research should analyse whether opinions differ depending on a
scientist’s expertise regarding a policy issue47 and public support for the policy in question48–50.
Future studies should also examine whether normative perceptions of science in policymaking
shift when specific scientific disciplines or policy issues are mentioned in real-world settings.
A majority of the public want scientists to prioritise research on public health and solving
energy problems. Yet, most people believe that scientists are currently not tackling these issues
This is a preprint and is undergoing peer review.
16
sufficiently and think that defence and military technology are prioritised too much. As the
perceived benefits of science are strongly correlated with trust in scientists, greater consideration
of public research priorities in scientific research and public funding presents an important
avenue to increase trust.
Newspapers, opinion pieces, and books3 have spread narratives of low public trust in
scientists. However, such claims remain largely unsubstantiated by empirical evidence. Our
Many Labs study provides decision-makers, scientists, and the public with large-scale and open
public opinion data on trust in scientists that can help these stakeholders maintain, and
potentially increase, trust in scientists.
Methods
Overview
The data underlying the analyses were collected in a global, pre-tested, pre-registered, cross-
sectional online survey (N = 71,417 participants in k = 67 countries) between November 2022
and August 2023 as part of the TISP Many Labs project (“Trust in Science and Science-Related
Populism”). TISP is an international, multidisciplinary consortium of 239 researchers from 167
institutions across all continents. Researchers conducted surveys within 87 post-hoc weighted
quota samples in 67 countries, using the same questionnaire translated into 37 languages. In the
following, we will describe the procedures used to collect and analyse the data. Further details
are available in the SI.
Institutional Review Board (IRB) Approval
Our questionnaire was considered exempt from full IRB review from the Harvard University-
Area Committee on the Use of Human Subjects (protocol # IRB22-1046) in August 2022. A
modified IRB application was submitted and considered exempt from full IRB review by the
Harvard University-Area Committee on the Use of Human Subjects in November 2022 (protocol
# IRB22-1046). All authors have informed themselves whether IRB approval was required from
their institutions and obtained IRB approval if necessary.
This is a preprint and is undergoing peer review.
17
Pre-test
A pre-test with n = 401 was conducted in the United States in October 2022 to validate the
measures used in the questionnaire. Average completion time was 14 minutes. The questionnaire
was slightly modified and two questions were added to the survey after the pre-test. Data from
the pre-test was not included in the final analyses.
Questionnaire
In total, we measured 111 variables. No identifiable information was collected. The complete
questionnaire (in English) is available at the Open Science Framework. The core questionnaire
contained the components described in the following. Participants were presented with these
components in the order in which they are explained below, but the order of questions and items
of multi-item scales was randomised.
Consent form
Participants were asked to carefully read a consent form (approved under IRB protocol # IRB22-
1046), which included some general information about the study and the anonymity of the data.
Demographic data - Part I
Participants who consented to participating in the study were then asked to indicate their gender
(0 = Female, 1 = Male, 2 = Prefer to self-describe, 99 = prefer not to say), age, and level of
education (1 = Did not attend school, 2 = Primary education, 3 = Secondary education (e.g., high
school), 4 = Higher education (e.g., university degree or higher education diploma)).
Attention check I
The first attention check asked participants to write the number “213” into a comment box.
Participants who failed the attention check were redirected to the end of the survey and were not
remunerated. See SI for details on how many respondents failed this attention check in the
overall sample and across countries.
Definition of science and scientists
Participants were presented with a definition of science and scientists: When we say “science”,
we mean the understanding we have about the world from observation and testing. When we say
This is a preprint and is undergoing peer review.
18
“scientists”, we mean people who study nature, medicine, physics, economics, history, and
psychology, among other things. This definition was based on the Wellcome Global Monitor7.
We added it because in-depth interviews conducted by the Monitor7 suggested that including a
definition would improve the reliability of cross-country comparisons.
Media use and engagement with science
Participants were asked how often they had come across information about science in ten
different places (e.g., in news articles in printed newspapers or magazines) and how often they
had engaged with science in four different ways (e.g., have conversations with friends) in the last
12 months (1 = never, 7 = once or more per day).
Open-ended questions
Participants were randomly assigned to answer one of two open-ended questions. One question
asked participants to write who they think benefits the most from science and why, and the
second question asked about their opinion on what makes a scientist trustworthy.
Perceived benefits of science
Participants were asked how much they perceived that scientific research benefits people like
themselves in their country (1 = not at all, 5 = very strongly), and which geographic region
benefits the most and the least from the work that scientists do (1 = Africa, 2 = Asia, 3 =
Australia and Oceania, 4 = Europe, 5 = Latin America, 6 = North America).
Desired and perceived goals of science
Participants were asked what goals scientists should prioritize (four items, 1 = very low priority,
5 = very high priority), and how strongly they believed that science aims to tackle these goals (1
= not at all, 5 = very strongly).
Normative perceptions of science and society
Participants rated their agreement to six statements (e.g., scientists should be more involved in
the policy-making process) (1 = strongly disagree, 5 = strongly agree). Five of these statements
were taken from40.
This is a preprint and is undergoing peer review.
19
Willingness to be vulnerable to scientists
Participants’ willingness to be vulnerable to scientific guidance was assessed with three items (1
= not at all, 5 = very strongly). Willingness to be vulnerable has been conceptualized as a
behavioural trust measure as it reflects the ceding of authority51.
Trust in scientists
Trust in scientists was assessed with 12 questions on four different dimensions of trustworthiness
(i.e., competence, integrity, benevolence, openness) (1 = very [unqualified], 5 = very
[qualified]), based on Besley et al.51. Psychometric analyses (e.g., scale reliability, exploratory
and confirmatory factor analyses, measurement invariance tests) can be found in the SI.
Trust in scientific methods
Participants indicated their level of agreement on whether scientific research methods are the
best way to find out if something is true or false (1 = strongly disagree, 5 = strongly agree).
General trust in scientists
A single question taken from Funk et al.52 was used to measure participants’ level of confidence
in scientists (1 = no confidence at all, 5 = a great deal of confidence).
Outspokenness about science
Three items based on McKeever et al.53 were used to measure outspokenness about science (e.g.,
I will share my opinions about scientific issues, regardless of what others think of them) (1 =
strongly disagree, 5 = strongly agree).
Science-related populism
Science-related populist attitudes were assessed with the SciPop Scale18, which comprises eight
items (1 = strongly disagree, 5 = strongly agree).
Attention check II
In the second attention check, participants were instructed to select “strongly disagree” to a
question. Participants who did not select “strongly disagree” were redirected to the of the survey
This is a preprint and is undergoing peer review.
20
and were not remunerated. See SI for details on how many respondents failed this attention
check in the overall sample and across countries.
Social dominance orientation
To assess social dominance orientation, we used four items from Pratto et al.54 (1 = extremely
opposed, 10 = extremely favour).
Trust in climate scientists
Participants were asked how much they trust in scientists in their country who work on climate
change (1 = not at all, 5 = very strongly).
Emotions about climate change
Nine different emotions (e.g., helpless) were assessed (1 = not at all, 5 = very strongly).
Perception of government action on climate change
Participants were asked about their perception of government action on climate change with
seven items (1 = strongly disagree, 5 = strongly agree)55.
Support for climate policies
Participants indicated their support for five policies (1 = not at all, 3 = very much; 4 = not
applicable, which was recoded as missing).
Perceptions of extreme weather events
Participants indicated to what extent they believe that climate change has increased the impact of
six weather events over the last decades (1 = not at all, 5 = very much). They also indicated
whether they expected that climate change will increase the impact of these events in the future
(1 = not at all, 5 = very much).
Demographic data - Part II
Participants indicated their household’s yearly net income (in local currency), their political
orientation on a spectrum from liberal to conservative (1 = strongly liberal, 5 = strongly
conservative, 99 = I don’t know) and on a spectrum from left-leaning to right-leaning (1 =
This is a preprint and is undergoing peer review.
21
strongly left-leaning, 5 = strongly right-leaning, 99 = I don’t know), their religiosity (1 = not
religious at all, 5 = very strongly religious), and whether they live in a rural or urban area.
Collaborators were allowed to add questions at the end of the survey. Additional questions did
not have to be approved by the lead author.
Translations
The original English survey was translated into the local language where necessary. Translations
were done by native speakers who were familiar with the study background and, in many cases,
had expertise on survey research and/or the conceptual underpinning of the measures. Minor
linguistic adjustments were made to the survey if deemed necessary. Major changes in the
wording of the original survey instrument had to be approved by the project lead. In total, the
survey instrument was translated into 36 languages and dialects (see SI).
Preregistration
We submitted a comprehensive preregistration prior to the data collection to the Open Science
Framework (OSF). It included detailed descriptions of our research questions and hypotheses,
instruments, data collection, and analytical procedures. Please see SI for deviations from the
preregistration.
Power analysis
To determine our minimum target sample size, we ran simulation-based power analyses using
the R package simr (v1.0.7)56, which is designed to conduct power analyses for generalized
linear mixed models such as those we used in the main study (for detailed information see SI).
Procedure and final sample
Data were collected in surveys that used quotas for age (five bins: 20% 18-29 years, 20% 30-39
years, 20% 40-49 years, 20% 50-59 years, 20% 60 years and older) and gender (two bins: 50%
male, 50% female). Participants had to be 18 years of age or older and provide informed consent
to participate in the study. Data were collected between November 2022 and August 2023. See
fig. S20 for an overview of survey periods across countries. The median completion time was 18
minutes.
This is a preprint and is undergoing peer review.
22
The surveys were programmed in Qualtrics. Participants that completed the survey were
remunerated according to the market research company’s local rates. All data was collected via
online surveys, except for the Democratic Republic of Congo, where participants were
interviewed in face-to-face interviews and responses recorded in Qualtrics by the interviewers.
Collaborators were instructed to work with the market research company Bilendi & Respondi,
except for most African countries, where collaborators collected data with MSi. Convenience
samples were not accepted.
A total of n = 71,629 individuals from 87 samples across k = 67 countries completed the
survey (n = 71,417 after exclusion of duplicate respondents). See table S12 for an overview of all
included countries and valid sample sizes across countries (i.e., after exclusion of duplicate
respondents) and SI for detailed characteristics of the final sample and the representativeness of
surveyed countries by income and regions (Tables S10-11).
Exclusion of non-completes and data quality test
We excluded all respondents who did not complete the survey, because they cancelled
participation during the survey, were filtered as their gender × age quota was already full, or
because they did not pass one of the two attention checks. 4.24% of respondents who reached the
first attention check did not pass it (“Please write the number 213 into the comment box”).
24.42% of respondents who reached the second attention check did not pass it (“To show us that
you are still paying attention, please select ‘strongly disagree’”; see Table S14). We excluded all
respondents who managed to complete the survey more than once despite countermeasures (e.g.,
IP address checks). In total, we excluded 212 duplicate respondents (Table S15).
Outlier value removal
We removed extreme outlier values for age and household income: Age outliers were defined as
values smaller than 18 and bigger than 100. Income outliers were defined as values that were
smaller than zero, equal to zero, or outside 5 × the interquartile range of the log-transformed
income distribution within each country after exclusion of values smaller than zero or equal to
zero. This led to the removal of the age values of 8 respondents and the removal of the income
values of 2,454 respondents (1,362 respondents indicated income values equal to or smaller than
0; 1,092 respondents indicated income values outside 5 × the interquartile range of the log-
This is a preprint and is undergoing peer review.
23
transformed income distribution within each country after exclusion of values equal to or smaller
than 0).
Post-hoc weighting
We computed post-stratification weights with the R package survey (v4.2-1)57 to ensure that our
models would estimate parameters that are representative for target populations in terms of
gender, age, and education and have more precise standard errors (SEs). We used raking58 to
compute four kinds of weights, i.e. (A) post-stratification weights at country level, (B) sample
size weights for each country, (C) post-stratification weights for the complete sample, and (D)
rescaled post-stratification weights for multilevel analyses (see Table S16 and SI for more
information).
Scale reliability
Scales were combined into indices and psychometric properties were assessed for all indices (see
SI), including scale reliability (Cronbach’s Alpha and Omega) and cross-country measurement
invariances. Scale reliability was good for all scales (see Tables S17-18 and SI for details).
Analyses
Factors explaining trust in scientists
To investigate explanatory factors of trust in scientists and explore how their influence varies
across countries, we ran a block-wise linear multilevel regression analysis with the R package
lme4 (v1.1-34)59. The model included rescaled post-stratification weights60.
All independent variables in the first and second block were scaled by country means and
country SDs. All independent variables in the third block were scaled by grand means and grands
SDs.
We first tried to fit a model with random intercepts and random effects for all
independent variables. However, this model failed to converge with three negative eigenvalues
and also had a singular fit, i.e., some random effects correlations were close to -1/+1, and some
random effects variances were close to 0. This was likely because the random effects structure
was too complex. Therefore, we simplified the model as follows: To test global effects of the
This is a preprint and is undergoing peer review.
24
independent variables on trust in scientists, we fitted a model that contained random intercepts
across countries (but no random effects) and inspected fixed effects estimates. To investigate
how the influence of independent variables varies across countries, we fitted separate models,
each of which contained random intercepts across countries and random effects for one particular
independent variable. For example, we fitted a regression model with random intercepts across
countries and random effects for political orientation (but no random effects for all other
independent variables) to assess how the effect of political orientation on trust in scientists varies
across countries. This entire procedure was completely in line with our preregistration.
Before we fitted the multilevel models, we confirmed that they would fit the data better than
fixed-effects models. First, we inspected intra-class correlations for trust in scientists (ICC =
0.170). Second, we ran a likelihood-ratio tests: It showed that a random-intercept null model
explaining trust in scientists had significantly better fit than a fixed-effects null model (χ² =
5,990.4, p < .001).
Moreover, we tested for multicollinearity of independent variables for the most complex
model, i.e., after inclusion of all three blocks of independent variables (see table S24). All
variance inflation factors (VIF) were well below even a very conservative threshold value of 461.
Normative perceptions of science in policymaking
To examine whether the public demands that scientists should take an active role in society and
policymaking, we ran two analyses: First, we computed weighted probabilities of responses to
the five items measuring these perceptions. This analysis provided estimates that are
approximately representative with regards to gender, age, education, and country sample size.
Second, we tested explanatory factors of normative perceptions of science in policy-making and
society: We fitted a linear multilevel regression model with the R package lme4 (v1.1-34)59,
which explained the average agreement to the five individual items measuring those perceptions,
included the rescaled post-stratification weights, and contained trust in scientists, science-related
populist attitudes, and sociodemographic variables as independent variables, i.e. gender (binary;
1 = male), age (continuous), education (binary; 1 = tertiary education), annual household income
in US dollar (continuous, log-transformed), place of residence (binary; 1 = urban), right-leaning
political orientation (continuous), conservative political orientation (continuous), and religiosity.
All independent variables were scaled by country means and country SDs.
This is a preprint and is undergoing peer review.
25
We specified random intercepts across countries and random effects for trust in scientists
and science-related populist attitudes. Significance tests of regression estimates relied on the
Satterthwaite method62. Before we fitted the multilevel model, we confirmed that it would fit the
data better than a fixed-effects model. First, we inspected the intra-class correlation of the
normative perceptions index (ICC = 0.103). Second, we ran a likelihood-ratio test, which showed
that a random-intercept null model had significantly better fit than a fixed-effects null model
(χ² = 3756.0, p < .001). Moreover, we tested for multicollinearity of independent variables (see
table S25). All VIF values were well below even a very conservative threshold value of 461.
Perceived and desired priorities of scientific research
To explore desires that scientists should prioritize four specific goals (improving public health,
solving energy problems, reducing poverty, developing defences and military technology) as
well as perceptions that science actually tackles these goals, we ran three analyses: First, we
inspected weighted mean values of responses to the four items measuring priority desires as well
as weighted mean values of responses to the four items measuring perceptions that science
actually devote efforts to the four goals.
Second, we ran weighted paired-samples t-tests to analyse if mean values of desires and
perceptions differed significantly from each other. These analyses provided estimates that are
approximately representative with regards to gender, age, education, and country sample size.
Third, we tested explanatory factors of the discrepancy between the desire that scientists should
prioritize the four goals and perceptions that science actually tackles them. To do so, we ran four
block-wise linear multilevel regression analyses with the R package lme4 (v1.1-34)59. Each
model explained the discrepancy between desires that scientists should prioritize one of the four
goals and perceptions that science actually tackles them, with higher outcome variable values
indicating that perceptions are more likely to exceed desires and lower outcome variable values
indicating that perceptions are more likely to stay behind desires. All models included rescaled
post-stratification weights60.
For each of the four models, we specified random intercepts across countries and random
effects for trust in scientists and science-related populist attitudes. Significance tests of
regression estimates relied on the Satterthwaite method62. Before we fitted the multilevel models,
we confirmed that they would fit the data better than fixed-effects models. First, we inspected the
intra-class correlations of the four discrepancy scores (health: ICC = 0.113; energy: ICC = 0.080;
This is a preprint and is undergoing peer review.
26
poverty: ICC = 0.135; defence: ICC = 0.108). Second, we ran likelihood-ratio tests, which
showed that random-intercept null models had significantly better fit than fixed-effects null
models (health: χ² = 3230.2, p < .001; energy: χ² = 2255.0, p < .001; poverty: χ² = 4811.4,
p < .001; defence: χ² = 3646.5, p < .001). Moreover, we tested for multicollinearity of
independent variables for the most complex models (see table S26). All VIF values were well
below even a very conservative threshold value of 461.
This is a preprint and is undergoing peer review.
27
References and Notes:
1. Sturgis, P., Brunton-Smith, I. & Jackson, J. Trust in science, social consensus and vaccine
confidence. Nat. Hum. Behav. 5, 1528–1534 (2021).
2. Algan, Y., Cohen, D., Davoine, E., Foucault, M. & Stantcheva, S. Trust in scientists in times
of pandemic: Panel evidence from 12 countries. Proc. Natl. Acad. Sci. 118, e2108576118
(2021).
3. Nichols, T. M. The Death of Expertise: The Campaign against Established Knowledge and
Why it Matters. (Oxford University Press, 2017).
4. Cologna, V. & Siegrist, M. The role of trust for climate change mitigation and adaptation
behaviour: A meta-analysis. J. Environ. Psychol. 69, 101428 (2020).
5. Cologna, V., Hoogendoorn, G. & Brick, C. To strike or not to strike? an investigation of the
determinants of strike participation at the Fridays for Future climate strikes in Switzerland.
PLOS ONE 16, e0257296 (2021).
6. IPSOS. Ipsos Global Trustworthiness Monitor: Stability in an Unstable World. (2022).
7. Wellcome Global Monitor. Wellcome Global Monitor: How does the world feel about
science and health? https://wellcome.org/sites/default/files/wellcome-global-monitor-
2018.pdf (2018).
8. Wellcome Global Monitor. Wellcome Global Monitor: How Covid-19 affected people’s
lives and their views about science. https://cms.wellcome.org/sites/default/files/2021-
11/Wellcome-Global-Monitor-Covid.pdf (2020).
9. Mede, N. G. Legacy media as inhibitors and drivers of public reservations against science:
global survey evidence on the link between media use and anti-science attitudes. Humanit.
Soc. Sci. Commun. 9, 1–11 (2022).
10. Rutjens, B. T. et al. Science Skepticism Across 24 Countries. Soc. Psychol. Personal. Sci.
13, 102–117 (2022).
11. West, J. D. & Bergstrom, C. T. Misinformation in and about science. Proc. Natl. Acad. Sci.
118, e1912444117 (2021).
12. Roozenbeek, J. et al. Susceptibility to misinformation about COVID-19 around the world. R.
Soc. Open Sci. 7, 201199 (2020).
13. Scharff, D. P. et al. More than Tuskegee: Understanding Mistrust about Research
Participation. J. Health Care Poor Underserved 21, 879–897 (2010).
14. Hendriks, F., Kienhues, D. & Bromme, R. Replication crisis = trust crisis? The effect of
successful vs failed replications on laypeople’s trust in researchers and research. Public
Underst. Sci. 29, 270–288 (2020).
15. Rutjens, B. T. & Većkalov, B. Conspiracy beliefs and science rejection. Curr. Opin. Psychol.
46, 101392 (2022).
16. Douglas, K. M. Are Conspiracy Theories Harmless? Span. J. Psychol. 24, e13 (2021).
17. Mede, N. G. & Schäfer, M. S. Science-related populism: Conceptualizing populist demands
toward science. Public Underst. Sci. 29, 473–491 (2020).
This is a preprint and is undergoing peer review.
28
18. Mede, N. G., Schäfer, M. S. & Füchslin, T. The SciPop Scale for Measuring Science-Related
Populist Attitudes in Surveys: Development, Test, and Validation. Int. J. Public Opin. Res.
33, 273–293 (2021).
19. Funk, C., Tyson, A., Kennedy, B. & Johnson. Science and Scientists Held in High Esteem
Across Global Publics. Pew Research Center Science & Society
https://www.pewresearch.org/science/2020/09/29/science-and-scientists-held-in-high-
esteem-across-global-publics/ (2020).
20. Li, N. & Qian, Y. Polarization of public trust in scientists between 1978 and 2018: Insights
from a cross-decade comparison using interpretable machine learning. Polit. Life Sci. 41, 45–
54 (2022).
21. Azevedo, F. & Jost, J. T. The ideological basis of antiscientific attitudes: Effects of
authoritarianism, conservatism, religiosity, social dominance, and system justification.
Group Process. Intergroup Relat. 24, 518–549 (2021).
22. Chan, E. Are the religious suspicious of science? Investigating religiosity, religious context,
and orientations towards science. Public Underst. Sci. 27, 967–984 (2018).
23. Gil de Zúñiga, H., Ardèvol-Abreu, A., Diehl, T., Gómez Patiño, M. & Liu, J. Trust in
Institutional Actors across 22 Countries. Examining Political, Science, and Media Trust
Around the World. (2019) doi:10.4185/RLCS-2019-1329en.
24. Bromme, R., Mede, N. G., Thomm, E., Kremer, B. & Ziegler, R. An anchor in troubled
times: Trust in science before and within the COVID-19 pandemic. PLOS ONE 17,
e0262823 (2022).
25. Besley, J. C. & Tiffany, L. A. What are you assessing when you measure “trust” in scientists
with a direct measure? Public Underst. Sci. 09636625231161302 (2023)
doi:10.1177/09636625231161302.
26. Ghai, S., Forscher, P. S. & Chuan-Peng, H. The illusion of generalizability in one big team
science study. Preprint at https://doi.org/10.31234/osf.io/avcsp (2023).
27. Forscher, P. S. et al. The Benefits, Barriers, and Risks of Big-Team Science. Perspect.
Psychol. Sci. 18, 607–623 (2023).
28. Odeny, B. & Bosurgi, R. Time to end parachute science. PLoS Med. 19, e1004099 (2022).
29. Hendriks, F., Kienhues, D. & Bromme, R. Measuring Laypeople’s Trust in Experts in a
Digital Age: The Muenster Epistemic Trustworthiness Inventory (METI). PLOS ONE 10,
e0139309 (2015).
30. Besley, J. C., Lee, N. M. & Pressgrove, G. Reassessing the Variables Used to Measure
Public Perceptions of Scientists. Sci. Commun. 1075547020949547 (2020)
doi:10.1177/1075547020949547.
31. Putnick, D. L. & Bornstein, M. H. Measurement invariance conventions and reporting: The
state of the art and future directions for psychological research. Dev. Rev. 41, 71–90 (2016).
32. Achterberg, P., de Koster, W. & van der Waal, J. A science confidence gap: Education, trust
in scientific methods, and trust in scientific institutions in the United States, 2014. Public
Underst. Sci. 26, 704–720 (2017).
This is a preprint and is undergoing peer review.
29
33. Johnson, C., Thigpen, C. & Funk, C. On the Intersection of Science and Religion. Pew
Research Center’s Religion & Public Life Project
https://www.pewresearch.org/religion/2020/08/26/on-the-intersection-of-science-and-
religion/ (2020).
34. Lasco, G. & Curato, N. Medical populism. Soc. Sci. Med. 221, 1–8 (2019).
35. The Cultural Authority of Science: Comparing across Europe, Asia, Africa and the
Americas. (Routledge, 2018).
36. Clausen, B., Kraay, A. & Nyiri, Z. Corruption and Confidence in Public Institutions:
Evidence from a Global Survey. World Bank Econ. Rev. 25, 212–249 (2011).
37. Jong-sung, Y. & Khagram, S. A Comparative Study of Inequality and Corruption. Am.
Sociol. Rev. 70, 136–157 (2005).
38. Transparency International. 2022 Corruption Perceptions Index. Transparency.org
https://www.transparency.org/en/cpi/2022 (2022).
39. Sturgis, P. & Allum, N. Science in Society: Re-Evaluating the Deficit Model of Public
Attitudes. Public Underst. Sci. 13, 55–74 (2004).
40. Cologna, V., Knutti, R., Oreskes, N. & Siegrist, M. Majority of German citizens, US citizens
and climate scientists support policy advocacy by climate researchers and expect greater
political engagement. Environ. Res. Lett. 16, 024011 (2021).
41. Besley, J. C. The National Science Foundation’s science and technology survey and support
for science funding, 2006–2014. Public Underst. Sci. 27, 94–109 (2018).
42. Hines, L. A., Gribble, R., Wessely, S., Dandeker, C. & Fear, N. T. Are the Armed Forces
Understood and Supported by the Public? A View from the United Kingdom. Armed Forces
Soc. 41, 688–713 (2015).
43. Bubela, T. et al. Science communication reconsidered. Nat. Biotechnol. 27, 514–518 (2009).
44. Fiske, S. T. & Dupree, C. Gaining trust as well as respect in communicating to motivated
audiences about science topics. Proc. Natl. Acad. Sci. 111, 13593–13597 (2014).
45. Myers, T. A. et al. Predictors of trust in the general science and climate science research of
US federal agencies. Public Underst. Sci. 26, 843–860 (2017).
46. McCright, A. M., Dentzman, K., Charters, M. & Dietz, T. The influence of political ideology
on trust in science. Environ. Res. Lett. 8, 044029 (2013).
47. Oreskes, N. What Is the Social Responsibility of Climate Scientists? Daedalus 149, 33–45
(2020).
48. Cologna, V., Baumberger, C., Knutti, R., Oreskes, N. & Berthold, A. The Communication of
Value Judgements and its Effects on Climate Scientists’ Perceived Trustworthiness. Environ.
Commun. 16, 1094–1107 (2022).
49. Beall, L., Myers, T. A., Kotcher, J., Vraga, E. K. & Maibach, E. W. Controversy matters:
Impacts of topic and solution controversy on the perceived credibility of a scientist who
advocates. PLOS ONE 12, e0187511 (2017).
50. Kotcher, J., Myers, T. A., Vraga, E. K., Stenhouse, N. & Maibach, E. W. Does Engagement
in Advocacy Hurt the Credibility of Scientists? Results from a Randomized National Survey
Experiment. Environ. Commun. 11, 415–429 (2017).
This is a preprint and is undergoing peer review.
30
51. Besley, J. C., Lee, N. M. & Pressgrove, G. Reassessing the Variables Used to Measure
Public Perceptions of Scientists. Sci. Commun. 43, 3–32 (2021).
52. Funk, C., Hefferon, M., Kennedy, B. & Johnson, C. Trust and Mistrust in Americans’ Views
of Scientific Experts. Pew Research Center Science & Society
https://www.pewresearch.org/science/2019/08/02/trust-and-mistrust-in-americans-views-of-
scientific-experts/ (2019).
53. McKeever, R., McKeever, B. W. & Li, J.-Y. Speaking up Online: Exploring Hostile Media
Perception, Health Behavior, and Other Antecedents of Communication. Journal. Mass
Commun. Q. 94, 812–832 (2017).
54. Pratto, F. et al. Social dominance in context and in individuals: Contextual moderation of
robust effects of social dominance orientation in 15 languages and 20 countries. Soc.
Psychol. Personal. Sci. 4, 587–599 (2013).
55. Hickman, C. et al. Climate anxiety in children and young people and their beliefs about
government responses to climate change: a global survey. Lancet Planet. Health 5, e863–
e873 (2021).
56. Green, P. & MacLeod, C. J. SIMR: an R package for power analysis of generalized linear
mixed models by simulation. Methods Ecol. Evol. 7, 493–498 (2016).
57. Lumley, T. survey: Analysis of Complex Survey Samples. (2023).
58. Battaglia, M. P., Hoaglin, D. C. & Frankel, M. R. Practical Considerations in Raking Survey
Data. Surv. Pract. 2, (2009).
59. Bates, D. et al. lme4: Linear Mixed-Effects Models using ‘Eigen’ and S4. (2023).
60. Patil, I. et al. datawizard: Easy Data Wrangling and Statistical Transformations. (2023).
61. O’brien, R. M. A Caution Regarding Rules of Thumb for Variance Inflation Factors. Qual.
Quant. 41, 673–690 (2007).
62. Satterthwaite, F. E. Synthesis of variance. Psychometrika 6, 309–316 (1941).
i The term “country” in this article refers to both sovereign states and territories not recognized as
such.
Supplementary Information is available for this paper.
Acknowledgments: We thank Damiano Lombardi (University of Zurich) for managing the
author list and author contributions.
Author contributions:
Conceptualization: VC, NM, SB, JB, CB, MJ, EWM, SM, NO, MS, SVDL
Data curation: NM
Formal analysis: NM
This is a preprint and is undergoing peer review.
31
Methodology: VC, NM, SB, JB, CB, MJ, EWM, SM, NO, MS, SVDL
Project admin.: VC
Software: VC, NM
Supervision: VC, NM
Validation: VC, NM
Visualization: VC, NM
Investigation: VC, NM, NIAA, SA, NAS, BA, IA, EA, AA, MA, MA, RMA, RA, TA,
DA, FA, AB, RB, KB, EB, AYB, OB, KB, AB, OB, PCA, FC, ACV, TC, RKC, SC, GC,
SDP, RD, SD, CDC, LDS, KD, SD, KMD, CD, DD, MD, UKHE, TWE, MF, AFB, ZF,
XF, CF, CF, MF, SF, HF, JF, MF, SF, AG, PGV, MGV, WG, OG, OG, TG, JG, EG,
MG, CGB, HG, DG, GMG, LG, HH, LNH, PH, ACHM, AH, GH, MH, MH, NI, MI,
MTI, YJ, TJ, CAJ, SJ, DJ, MZK, JK, SK, JRK, MK, TKR, OK, HK, AK, LK, EK, LK,
AK, LSK, PK, SK, TK, AK, EAK, CL, AL, AL, JBL, ZL, NL, AL, GL, AL, ALO, CLV,
NML, CHL, KLT, MDM, SM, HM, JM, TLM, JM, PM, FMR, MM, IM, ZM, JN, EJN,
JPN, FLNC, DN, TO, JPH, MP, PP, PPS, MPC, MP, YP, ARP, MP, KP, MP, JP, DP,
AP, KP, EP, KP, DMQ, PR, AR, FGR, CRS, GR, JR, SR, JPR, RMR, IR, OS, RRS, PS,
BS, AS, JSN, ES, NS, LS, BS, OS, SKS, GS, SS, BS, ES, MT, CTE, CTE, BT, AKT,
RT, DTF, MT, OMU, ICU, JVN, CV, SV, IV, AVB, IW, IW, MW, TW, MW, FW,
ADW, ZX, JX, EZP, AZ, RAZ
Resources: NIAA, SA, NAS, BA, IA, EA, AA, MA, MA, MA, RMA, RA, TA, DA, FA,
AB, RB, KB, EB, AYB, OB, KB, AB, OB, PCA, FC, ACV, TC, RKC, SC, GC, SDP,
RD, SD, CDC, LDS, KD, SD, KMD, CD, DD, MD, UKHE, TWE, MF, AFB, ZF, XF,
CF, CF, MF, SF, HF, JF, MF, SF, AG, PGV, MGV, WG, OG, OG, TG, JG, EG, MG,
CGB, HG, DG, GMG, LG, HH, LNH, PH, ACHM, AH, GH, MH, MH, NI, MI, MTI, YJ,
TJ, CAJ, SJ, DJ, MZK, JK, SK, JRK, MK, TKR, OK, HK, AK, LK, EK, LK, AK, LSK,
PK, SK, TK, AK, EAK, CL, AL, AL, JBL, ZL, NL, AL, GL, AL, ALO, CLV, NML,
CHL, KLT, MDM, SM, HM, JM, TLM, JM, PM, FMR, MM, IM, ZM, JN, EJN, JPN,
FLNC, DN, TO, JPH, MP, PP, PPS, MPC, MP, YP, ARP, MP, KP, MP, JP, DP, AP, KP,
EP, KP, DMQ, PR, AR, FGR, CRS, GR, JR, SR, JPR, RMR, IR, OS, RRS, PS, BS, AS,
JSN, ES, NS, LS, BS, OS, SKS, SS, BS, ES, MT, CTE, CTE, BT, AKT, RT, DTF, MT,
OMU, ICU, JVN, CV, SV, IV, AVB, IW, IW, MW, TW, MW, FW, ADW, ZX, JX, EZP,
AZ, RAZ
Funding acquisition: VC, NO, MS, JB, EM, SB, CB, BA, IA, EA, MA, MA, RMA, DA,
AB, RB, EB, CB, AYB, RB, OB, MB, AB, KB, AB, OB, PCA, FC, ACV, TC, SC, GC,
RD, SD, CDC, LDS, KD, SD, KMD, CD, DD, MD, UKHE, ME, BE, TWE, MF, AFB,
XF, CF, CF, MF, SF, HF, JF, MF, SF, AG, MGV, WG, OG, OG, TG, JG, EG, MG,
CGB, HG, DG, GMG, LG, HH, LNH, PH, ACHM, AH, GH, MH, MH, NI, MI, CAJ, SJ,
DJ, MZK, JK, SK, JRK, TKR, OK, HK, TK, AK, LK, EK, LK, LK, AK, JK, LSK, PK,
SK, AK, EAK, CL, AL, AL, JBL, ZL, NL, AL, GL, AL, ALO, CLV, NML, CHL, KLT,
MDM, SM, RM, HM, JM, TLM, JM, PM, FMR, MM, EJN, JPN, TO, TO, JPH, MP, PP,
PPS, MPC, MP, YP, ARP, MP, CRP, KP, JP, DP, AP, EP, KP, PR, AR, FGR, CRS, JPR,
JR, SR, JPR, RMR, IR, OS, RRS, PS, SS, BS, AS, JSN, ES, JS, NS, LS, BS, OS, SKS,
GS, SS, ES, MT, CTE, CTE, BT, RT, DTF, MT, MT, OMU, ICU, JVN, CV, SV, IV,
AVB, IW, IW, MW, TW, MW, FW, ADW, ZX, JX, EZP, AZ, RAZ
This is a preprint and is undergoing peer review.
32
Writing-orig. draft: VC
Writing – review & editing: VC, NM, SB, JB, CB, MJ, EWM, SM, NO, MS, SVDL,
NIAA, SA, NAS, BA, IA, EA, AA, MA, MA, MA, RMA, RA, TA, PA, DA, FA, AB,
RB, KB, EB, CB, AYB, RB, OB, MB, AB, KB, AB, OB, PCA, FC, ACV, TC, RKC, SC,
GC, SDP, RD, SD, CDC, LDS, KD, SD, KMD, CD, DD, MD, UKHE, ME, BE, TWE,
MF, AFB, ZF, XF, CF, CF, MF, SF, HF, JF, MF, SF, AG, PGV, MGV, WG, OG, OG,
TG, JG, EG, MG, CGB, HG, DG, GMG, LG, HH, LNH, PH, ACHM, AH, GH, MH,
MH, NI, MI, MTI, YJ, TJ, CAJ, SJ, DJ, MZK, JK, SK, JRK, MK, TKR, OK, HK, AK,
LK, EK, LK, LK, AK, JK, LSK, PK, SK, TK, AK, EAK, CL, AL, AL, OLC, JBL, ZL,
NL, AL, GL, AL, ALO, CLV, NML, CHL, KLT, MDM, SM, RM, HM, JM, TLM, JM,
PM, FMR, MM, IM, ZM, JN, EJN, JPN, FLNC, DN, TO, TO, JPH, MP, PP, PPS, MPC,
MP, YP, ARP, MP, CRP, KP, MP, JP, DP, AP, KP, EP, KP, DMQ, PR, AR, FGR, CRS,
GR, JPR, JR, SR, JPR, RMR, IR, OS, RRS, PS, SS, BS, AS, JSN, ES, JS, NS, LS, BS,
OS, SKS, GS, SS, BS, ES, MT, CTE, CTE, BT, AKT, RT, DTF, MT, MT, OMU, ICU,
JVN, CV, SV, IV, MV, AVB, IW, IW, MW, TW, MW, FW, ADW, ZX, JX, EZP, AZ,
RAZ
Competing interests: Authors declare that they have no competing interests.
Funding sources: Anonymized for this preprint. Available upon request.
Materials & Correspondence. Correspondence and requests for materials should be addressed
to Viktoria Cologna (v.cologna@ikmz.uzh.ch).
