The Communication of Value Judgements and its Effects on Climate Scientists’ Perceived Trustworthiness

ABSTRACT

Scientists are called upon by policymakers to provide recommendations on how to address climate change. It has been argued that as policy advisors, scientists can legitimately make instrumental value judgements (recommendations based on defined policy goals), but not categorical value judgements (challenge and/or redefine established policy goals), and that to do otherwise is to overstep in ways that may threaten their perceived trustworthiness. However, whether these types of value judgements affect public trust in scientists remains largely untested. We conducted two studies (N₁ = 367, N₂ = 819) to investigate public perceptions of trustworthiness of a climate scientist expressing either an instrumental or a categorical value judgement. We found no difference in perceived trustworthiness between the two conditions. However, trustworthiness perceptions in both studies depended on individuals’ support for the policy recommended by the scientist. Our findings suggest that climate scientists should not fear for their overall perceived trustworthiness when making categorical value judgments if their opinions are supported by the majority of the public.

KEYWORDS:

• Climate change
• science communication
• value-free ideal
• credibility
• advocacy
• trust

1. Introduction

Policy-makers are often epistemically dependent upon expert knowledge to implement effective policies (Almassi, 2012; Oppenheimer et al., 2019). In this process, scientists may be expected to act as “honest brokers” of policy alternatives (Pielke, 2007), i.e. to clarify and expand possible options for policy action without (necessarily) advocating for any particular policy. This is the role that the Intergovernmental Panel on Climate Change (IPCC) envisions for scientists qua policy advisors: to be neutral, policy- relevant, but not policy-prescriptive. This view rests on an ideal of value-free science: the idea that non-epistemic values (e.g. ethical and social values) can legitimately play a role in the choice of research questions, but not in the evaluation and acceptance of evidence and hypotheses (Douglas, 2009; Proctor, 1991; Raman, 1993). In the case of scientists qua policy advisors, the value-free ideal recommends that the influence of scientists’ own moral and political values should be minimized when advising policymakers and communicating policy recommendations (Gundersen, 2020).

However, a recent qualitative study showed that most IPCC scientists perceive themselves to have the moral responsibility to act as more than just policy-neutral advisors (Gundersen, 2020). Yet, the same scientists view themselves as to adhering to the value-free ideal of science, convinced that their public credibility depends on their perceived neutrality and impartiality (Gundersen, 2020). This discrepancy between scientists’ expectations of themselves and their reticence to act on those expectations has been described as a perception-engagement gap (Cologna et al., 2021). Research suggests that the majority of the public in Germany and the US support policy advocacy by climate change researchers (Cologna et al., 2021). Given this, are climate scientists right to fear for their perceived trustworthiness when making value judgements in policy recommendation? Before reviewing the literature on scientists’ advocacy and its effects on scientists’ perceived trustworthiness, we will first describe the two types of value judgements that are relevant in this research.

1.1. Instrumental and categorical value judgements

A value judgement is a judgement based upon a particular set of values or on a particular value system. In 1965, Hempel distinguished between two types of value judgements that we deem to be relevant in environmental communication: instrumental or relative value judgements vs. categorical or absolute value judgements. An instrumental value judgement is a conditional or relative statement, which asserts that “a certain kind of action, M, is good (or that it is better than a given alternative M₁) if a specified goal G is to be attained” (Hempel, 1965, p. 84). In climate science and policy, an instrumental value judgement would be to say that a carbon tax (policy P), is an effective policy if a reduction in CO₂ emissions (goal G) is desired. These statements can also take the form of probabilistic claims, such as “the introduction of policy P has a certain probability X of achieving goal G”. As these examples show, an instrumental value judgement can be formulated without using terms of moral discourse such as “good”, “better” or “ought to”. Such a judgement simply expresses a means-end-relation. More precisely, it claims that something is a (definitively or probabilistically) effective (or necessary or sufficient) means to achieve a given end. Therefore, instrumental value judgements are empirical assertions susceptible to scientific testing (Hempel, 1965).

Instrumental value judgements rely on existing norms or goals that have been formulated by policy makers or society at large. Schurz (2013) argues that in order not to undermine their credibility, scientists must explicitly relativize their policy recommendations according to existing societal norms and goals. In the case of climate change, the target ratified under the Paris Agreement to limit global warming to well below 2°C would be such a normative goal, as it was agreed upon by policymakers.

If, in contrast, scientists were to question policy goals, rather than merely recommending policies as means to achieve democratically agreed-upon goals, this would constitute a categorical value judgement. A categorical judgement can be defined as a “judgement of value to the effect that a certain state of affairs (which may have been proposed as a goal or end) is good, or that it is better than some specified alternative” (Hempel, 1965, p. 85). For example, if a scientist questions the internationally agreed upon 2°C target because she judges the consequences of such warming to be too severe or not severe enough to take action accordingly, this would constitute a categorical value judgement. In this case, a scientist expresses preferences for goals that might be legitimate, but have not been democratically agreed upon.

Categorical value judgements are not susceptible to scientific test and confirmation as they express normative rather than descriptive assertions, although certain normative choices may well be democratically agreed upon (e.g. the principle of “common but differentiated responsibility” in the UNFCCC (UNFCCC, 1992)). Because the use of categorical value judgements contrasts with the value-free ideal of science, their use in policy recommendation settings is often considered illegitimate, and assumed to be damaging to public trust in science. As Max Weber is said to have put it: “Science is like a map: it can tell us how to get to a given place, but it cannot tell us where to go” (Hempel, 1965, p. 86). To sum up, categorical value judgements claim that a certain goal is good or better than some other goal, and are thus normative assertions that are not capable of scientific testing. Instrumental value judgements, on the other hand, claim that a certain mean is effective (or necessary or sufficient) to achieve a given goal, and are thus empirical assertions that are (at least in principle) capable of scientific testing.

When scientists invoke values in the context of policy recommendation, the word “advocacy” is often employed, usually with negative connotation. An advocate can decide to support a certain policy based on an instrumental value judgement (“policy P is the best policy to achieve goal G and should therefore be implemented”) or a categorical value judgement (“Goal G₂ rather than established policy goal G₁ should be pursued, with policy P₂ being the best policy to achieve goal G₂. Thus, policy P₂ should be implemented”). Previous work on advocacy and value judgements has generally elided this difference or focused on the communication of instrumental value judgements. This presents a significant gap in the literature, as there is a consensus on the legitimacy of instrumental value judgements by scientists qua policy advisors, while opinions strongly diverge on the legitimacy of categorical value judgements. Our hypothesis is that the perceived trustworthiness of an advocating scientist depends on the kind of value judgement underlying the statement as well as the preexisting support for the policy that is being advocated for; that is what we test here.

1.2. Values and trust in science

One of the main arguments presented in favor of value-free science is that the use of non-epistemic values in categorical value judgements undermines the foundation for public trust in science, which is asserted to be based on science being objective, interpreted as being free from value-influence (Betz, 2013; Schroeder, 2019; Schurz, 2013). Against that, many historians and philosophers argue that the value-free ideal of science is untenable. Proctor (1991) has argued that the ideal of value-free science is not itself value-free, but was promulgated in the late 19th and early twentieth century for what were often in fact political purposes. Others have argued that science simply is not and never has been value-free; the value-free ideal does not offer an empirically adequate account of scientific practice (Jasanoff, 1990; Kitcher, 2011; Kourany, 2008; Oreskes, 2019). Going further, philosophers of science have argued for a positive role for value judgements, because they can lead to interpretive options that might otherwise be missed (Douglas, 2000, 2009; Elliott, 2017; Lloyd, 2015; Longino, 1990; Steele, 2012).

In climate science, several examples demonstrate that while scientists can try to be transparent about their value judgements, they cannot escape making value judgements when advising policymakers, or even in the scientific work itself (see for example Brysse et al., 2013; Steele, 2012; Winsberg et al., 2020). Some authors have argued that “the willingness to make important judgments more explicit and to subject them to critical scrutiny is a hallmark of objectivity” (Elliott, 2017, p. 15) and that transparent communication of value judgements may act to preserve trust in science (Elliott & Resnik, 2014; Oreskes, 2019).

Some studies have attempted to test these positions by analysing whether policy recommendation and advocacy by climate scientists influence scientists’ credibility/trustworthiness as perceived by the public (Beall et al., 2017; Cologna et al., 2021; Kotcher et al., 2017; Palm et al., 2020). However, these studies did not distinguish between instrumental and categorical value judgements. We are only aware of one study that specifically investigated the effects of communicating categorical value judgements on scientists’ perceived trustworthiness (Elliott et al., 2017). These authors examined the impact of three variables: 1) whether or not a public-health scientist expressed categorical value judgements (promoting economic growth or protecting public health), 2) whether the scientist's conclusion appeared contrary to or consistent with the scientist's values, and 3) whether subjects’ values aligned with the scientist's values. Even though the authors do not explicitly describe these value judgements as categorical value judgements, we understand this to be the first study to specifically analyse the effect of categorical value judgements on public trust in the communicating scientist. Elliott and colleagues (2017) conclude that trust may be contingent on the value congruence between the scientist and the reader.

The idea that value similarity affects levels of trust is supported by other studies and across different domains (Earle et al., 2007; Poortinga & Pidgeon, 2006; Siegrist et al., 2000). These studies find that we tend to trust people whose values or intentions are similar to ours, and to assign greater credibility to policy advocates who share our values or intentions (Kahan et al., 2008). In this context, Kahan and colleagues (2008, 2010) speak of a “cultural credibility heuristic”: we tend to rely on trusted actors to determine which information to believe, and are in turn inclined to trust those whose values we share. Cologna et al. (2021) found that the perceived legitimacy of advocacy differs depending on the policy in question, and hypothesized that advocacy will likely be perceived as more legitimate when scientists advocate for policies with higher public support. In this sense, support for the policy recommended by a scientist can be conceptualized as a proxy for similarity of intentions/values. When determining whether an unknown scientist is trustworthy, people might therefore rely on the perceived similarity of values based on the policies recommended by the scientist. In the current study, we therefore assess how participants’ preexisting support for the policy called for by scientists influences their perceptions of the scientist’s trustworthiness.

Understanding how the communication of different value judgements impacts scientists’ perceived trustworthiness is important, especially as climate change researchers are increasingly signing statements to support youth climate strikes (Hagedorn et al., 2019), issuing warnings of a climate emergency (Hagedorn et al., 2019; Ripple et al., 2020; The Guardian, 2019; Warren, 2019) and endorsing protests calling for political action on climate change (Green, 2019; Reuters, 2019). Understanding of how different types of values in science communication affect climate change researchers’ perceived trustworthiness can help them to communicate more effectively with the public, while maintaining, and possibly increasing, public trust. This is especially important, as research shows that the more the public trusts climate scientists, the more they are likely to accept the reality of climate change (Hornsey et al., 2016), be concerned about it (Malka et al., 2009) and undertake climate-friendly behavior and support for climate policies (Cologna et al., 2022; Cologna & Siegrist, 2020). Further, research shows that individuals rely on source credibility when judging the validity of scientific claims (Scharrer et al., 2019) with source credibility predicting plausibility perceptions of scientific statements about climate change (Lombardi et al., 2014).

Different definitions and conceptualisations of trust have been proposed in the literature. One of the most cited definitions describes trust as “a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behaviour of another” (Rousseau et al., 1998, p. 395). Trust is therefore described as a behavior that entails making oneself vulnerable to others. In our case, trustworthiness is understood as the perception that an actor is worthy of our trust, and less as an actual behavior that implies a certain degree of vulnerability. Relevant for this study is the concept of “epistemic trustworthiness” which describes experts’ characteristics that recipients depend on when judging whether to defer to experts due to their limited resources (Hendriks et al., 2015). Some authors have operationalized trustworthiness by focusing on character-based elements (e.g. honesty, integrity) (Cologna et al., 2021; Renn & Levine, 1991), while others have used a combination of both competence- and character-based elements (Fiske & Dupree, 2014; Hendriks et al., 2015). In Study 1, we use a scale that focuses on character-based elements (Cologna et al., 2021), while in Study 2 we use a scale that additionally accounts for competence-based elements (Hendriks et al., 2015) – an element of trustworthiness that we deemed important in the context of this study. Indeed, recent research has shown that character- and competence-based factors can clearly be distinguished in a factor analysis as two dimensions of trustworthiness (Besley et al., 2021).

Our paper advances previous efforts by investigating the following research questions: 1) Does the communication of a categorical value judgement, as opposed to an instrumental value judgement, influence the perceived trustworthiness of the communicating climate scientist? 2) Does support for the recommended policy influence the perceived trustworthiness of a scientist expressing instrumental and categorical value judgements? We also consider how other factors (e.g. age, gender, political orientation) influence the perceived trustworthiness of scientists communicating value judgements. We test the hypothesis that scientists expressing categorical value judgements will be perceived as less trustworthy than scientists expressing instrumental value judgements. We also test the hypothesis that the perceived trustworthiness of scientists expressing value judgements will depend on the public’s support for the policy being recommended or advocated for.

2. Study 1

2.1. Methods

2.1.1. Sample

To recruit participants from the German-speaking part of Switzerland, we invited N = 600 participants through the use of the ETH Zurich Consumer Behavior Internetpanel. Overall, N = 378 participants completed our survey, a response rate of 62%. Surveyed individuals participated on a voluntary basis and were not remunerated. We excluded participants (n = 11) whose mean values regarding the dependent variable in each condition deviated by 2 or more SDs from the mean. Our final sample thus consisted of N = 367 participants. The participants in our survey had an average age of 65 years (SD = 12) which is higher than the average age in Switzerland of 43 (BFS, 2020). 38% of participants identified as female, so women were underrepresented in our survey. The average level of education in our sample was also higher than the average level of education in Switzerland. The unrepresentativeness of our sample is discussed in the limitation section. This study was approved by the ETH Zurich Ethics Commission: EK 2020-N-125.

2.1.2. Questionnaire

Participants were asked to provide socio-demographic information regarding their gender, year of birth, education and political orientation. Their support for five different climate-related policies, including the support for a stricter CO₂ tax, was assessed on a four-point Likert scale (1 = do not support at all to 4 = fully support). As the scientists in our scenarios recommended the adoption of stricter CO₂ taxes, we specifically measured support for this strongly debated policy.

Participants were randomly assigned to the two conditions and asked to read a tweet carefully. The tweets were created in the name of a fictional Prof. Keller (a common Swiss surname).

After reading this statement, one group was shown a tweet that contained the following instrumental value judgement:

My results show that various climate policies can be effective at mitigating climate change. One such effective policy is the introduction of higher CO₂ taxes.

The second group was shown a tweet that contained the following categorical value judgement:

Given my research findings, I am convinced that the internationally agreed upon climate goals do not suffice to avert the worst risks of climate change. The government should therefore introduce way stronger measures than the internationally agreed upon measures to mitigate the risks of climate change. I, therefore, urge the government to increase CO₂ taxes accordingly.

To assess Professor Keller’s perceived trustworthiness, we used four items that asked whether Prof. Keller is trustworthy, honest in reporting research findings, objective when discussing climate policies with politicians and the media, and acting in the interest of society. These items were rated on a six point Likert scale from 1 = strongly disagree, to 6 = strongly agree (Cologna et al., 2021). We added a fifth item asking whether Prof. Keller is pursuing a secret agenda. However, as the addition of the fifth item decreased the reliability of the scales, we decided to remove the fifth item; this led to a strong reliability of the four-item scale in both the instrumental (Cronbach’s alpha = .91) and the categorical value judgement condition (Cronbach’s alpha = .91).

Participants were then asked whether they believe that climate change is happening and if so, whether it is human-caused (Leiserowitz et al., 2019). Climate change concern was assessed on a six-point Likert scale (1 = strongly disagree to 6 = strongly agree) with four items from Shi et al. (2016) (Cronbach’s alpha = .92). Finally, participants were asked how strong they perceive the scientific consensus to be on both the existence of climate change and on climate change being mainly human-caused (Leiserowitz et al., 2019).

2.1.3. Data analysis

We conducted an independent t-test to investigate if there is a difference between the two conditions regarding the overall trustworthiness score. To determine the effects of the between subjects’ factors condition (instrumental vs categorical) together with the participants’ level of support for the policy recommend by the scientist on the dependent variable level of perceived trustworthiness, we used a linear regression analysis and standardized our interaction term as recommended in the literature (Cohen et al., 2002). We also analysed the effects of our predictors (i.e. support for CO₂ taxes, climate change concern, gender, political orientation, age) with further multiple linear regression analyses to examine how perceived trustworthiness is affected by these variables for both conditions.

2.2. Results

Against our hypothesis, we found no overall difference in the level of perceived trustworthiness between the instrumental (M = 4.35, SD = 0.94) and the categorical (M = 4.22, SD = 1.03) value judgement condition (t(365) = -1.23, d = -.13, p = .22), meaning that a climate scientist expressing a simple means-end relationship in a climate policy context was not perceived to be more or less trustworthy than a climate scientist who questioned existing policy goals and urged for a specific policy. Inspecting individual items revealed only one significant, albeit small, difference: the scientist in the categorical value judgement condition who questioned existing policy goals was perceived to be less objective than the scientist in the instrumental value judgement condition who recommended a climate policy as a means to mitigate climate change (t(365) = -1.99, d = -.21, p < .05).

When considering both the condition and the level of support for a CO₂ tax, we find that support for the policy recommended by the scientist strongly predicts perceived trustworthiness ($\text{β}$=  0.62, SE = 0.05, p < .001) while the type of condition (reference group: categorical value judgement) did not ($\text{β}$=  0.03, SE = 0.08, p = .50). Thus, irrespective of the condition, support for the policy recommended by the scientist always predicts higher levels of perceived trustworthiness. Adding an interaction term to the model (policy support x condition) did not result in a significant effect on perceived trustworthiness levels ($\text{β}$=  −0.09, SE = 0.08, p = .10). Overall, the effect of different value judgements on scientists’ perceived trustworthiness seems to be dependent on the level of support for the policy the scientist is recommending, confirming our second hypothesis.

Multiple regression analyses showed that climate change concern was a significant predictor of perceived trustworthiness (Table 1). In both groups, the more concerned participants were about climate change, the more trustworthy they perceived scientists to be. Further, a left-wing political orientation significantly predicted perceived trustworthiness of the scientist. Older participants perceived scientists communicating instrumental value-judgements to be more trustworthy than younger participants did, but age was not a significant predictor of the level of credibility in the categorical value judgement condition. It should be noted that our models explain a considerable amount of variance, with the instrumental value judgement model explaining 46% of variance and the categorical value judgement model explaining 55% of variance.

Table 1. Predictors of perceived trustworthiness for the instrumental and the categorical value judgement conditions.

	Perceived trustworthiness
	Instrumental value judgement				Categorical value judgement
Predictors	B	95% CI	$\text{β}$	p	B	95% CI	$\text{β}$	p
(Intercept)	0.88	[−0.24, 2.00]		.121	1.46	[0.50, 2.41]		.003
Support CO2 tax	0.31	[0.14, 0.48]	0.28	<.001	0.39	[0.23, 0.54]	0.33	< .001
Climate change concern	0.32	[0.16, 0.48]	0.29	<.001	0.36	[0.22, 0.49]	0.34	< .001
Male	0.18	[−0.05, 0.40]	0.09	.119	−0.10	[−0.32, 0.11]	−0.05	.341
Political orientation	−0.01	[−0.02, −0.00]	−0.19	.012	−0.01	[−0.02, −0.00]	−0.19	.002
Age	0.02	[0.01, 0.02]	0.22	< .001	0.01	[−0.00, 0.02]	0.09	.066
N = 173, R^2 adjusted = 45.8%			N = 194, R^2 adjusted = 54.6%

Note. Significant values are printed in bold. The following measures are presented in this table: Support for a CO₂ tax (1 = do not support at all to 4 = fully support); Climate change concern ((1 = strongly disagree to 6 = strongly agree); Political orientation (1 = strongly left-wing to 100 = strongly right-wing).

2.3. Discussion and limitations of Study 1

Against our first hypothesis, we did not find an overall difference in the perceived trustworthiness between the scientist expressing an instrumental value judgement versus a categorical value judgement. Only the scientists’ perceived level of objectivity differed between conditions, with the scientist in the instrumental value judgement condition being perceived as more objective than the scientist in the categorical value judgement condition. In line with our second hypothesis, we found that perceived trustworthiness was strongly predicted by respondents’ support for the policy recommended by the scientists.

Our study has several limitations that may have influenced the results. First, our sample was not representative of the Swiss population, with male, older and more educated participants being overrepresented. Second, our study was very likely underpowered to detect a significant interaction effect. Third, the two tweets differed in length, with the tweet expressing a categorical value judgement being 28 words longer than the tweet expressing an instrumental value judgement. This might have strengthened differences between participants with lower and higher policy support, as participants with lower policy support could have been less willing to engage with a longer text. Fourth, the language used in the two tweets differed: the first tweet used passive language when recommending stricter CO₂ taxes, while the second text used active language. Further, the advocacy statement was more pronounced in the categorical value judgement tweet compared to the instrumental value judgement tweet. This might have introduced some potential confounds. For these reasons, we decided to conduct a second well-powered study with a representative sample to see if we could confirm the results found in Study 1.

3. Study 2

3.1. Methods

3.1.1. Sample

To determine the sample size needed to have enough power (85%) to detect the hypothesized main and interaction effect, we ran a simulation-based power analysis: we simulated data based on the means and correlations we found in Study 1. Our power analysis indicated that with a power of .85 and a threshold of p < .05, we would need a sample size of approximately N = 810 to detect a significant interaction effect of condition x policy support on perceived trustworthiness evaluations. With this sample size, our study would also be sufficiently powered to detect main effects of treatment (0.99) and policy support (1.00). We preregistered our study and attached our simulation-based power analysis: https://osf.io/c2agd. As per preregistration, we aimed to recruit N = 880 participants in order to have a large enough sample should we have to exclude participants. Through the market research company Respondi, we recruited N = 883 German speaking respondents from Switzerland and cross-quoted participants for gender and age. We excluded participants (n = 64) whose mean values of the dependent trustworthiness variables differed by equally or more than 2 SDs from the mean, reaching a final sample of N = 819. In contrast to Study 1, our sample in Study 2 was representative in terms of gender, with 53% identifying as female, 47% identifying as male and 0.2% selecting the option “diverse”. The sample was also representative in terms of education and age, with an average of 45 years (SD = 14), close to the average Swiss age of 43 (BFS, 2020). This study was approved by the ETH Zurich Ethics Commission: 2021-N-201.

3.1.2. Questionnaire

After filling out demographic details, participants were asked to state to what extent they agreed with the introduction of four different climate policies including stricter CO₂ taxes and higher subsidizes for renewable energies. We chose these two policies as Cologna et al. (2021) have shown support for advocacy of these two policies to differ. Given that in Study 1 we found policy support for a CO₂ tax to predict perceived trustworthiness of a scientist recommending a higher CO₂ tax, we wanted to investigate whether this would also hold true for other policies, such as subsidies for renewable energy. To capture additional variance, we changed the scale from a four-point scale used in Study 1 to a six-point scale (1 = strongly disagree to 6 = strongly agree).

To assess the perceived trustworthiness of a scientist expressing an instrumental or a categorical value judgement, we displayed tweets containing these value judgements (as in Study 1). However, instead of assigning participants to one of the two conditions (instrumental versus categorical value judgment condition), we used a within-subjects design where participants were presented with both tweets containing instrumental value judgements and categorical value judgements.

Participants were told that they would be presented with several tweets by famous sportsmen and renowned scientists and asked to carefully read all the tweets before being asked to assess the perceived trustworthiness of the person who wrote the tweets. Overall, participants saw six tweets in randomized order, three tweets by Swiss sportsmen (one of which was always shown first), and four tweets (of similar length) by renowned scientists displaying: 1) an instrumental value judgement calling for higher CO₂ taxes, 2) an instrumental value judgement calling for higher subsidies for renewable energy, 3) a categorical value judgement that questioned the effectiveness of current climate policies and called for higher CO₂ taxes and 4) a categorical value judgement that questioned the effectiveness of current climate policies and called for higher subsidies for renewable energy. Names and pictures of scientists in the tweets were blurred. At the end of the survey, participants were told that the tweets they had seen were not written by actual climate scientists, but had been created for the purpose of this study.

We decided to use a different scale to assess perceived trustworthiness than in Study 1, as we concluded that the scale used there (Cologna et al., 2021) misses an important element of perceived trustworthiness: perceived expertise. We, therefore, used a shortened version of the The Muenster Epistemic Trustworthiness Inventory (METI) to assess perceived trustworthiness (Hendriks et al., 2015). The METI was conceptualized to measure three important elements of trustworthiness: expertise, integrity and benevolence. Of each element, we used two items with the highest factors loadings in Hendriks et al. (2015)¹ and assessed them on a semantic differential from 1-7, for example “competent–incompetent”. We additionally added the item “objective-biased” as we observed differences in the level of objectivity between the two conditions in Study 1. Lastly, we measured climate change concern with the same measure as in Study 1 (Shi et al., 2016).

3.1.3. Data analysis

Given the within-subjects design of our study, we ran a multilevel linear model with the lme4 package in R (Bates et al., 2015). As preregistered, our model contained an interaction term (condition x policy support) with the variable policy support being standardized as recommended in the literature (Cohen et al., 2002). Given that perceptions of trustworthiness were nested within subjects, we allowed model intercepts to vary across subjects. We conducted multiple regression analyses to investigate predictors of trustworthiness perceptions for both conditions.

3.2. Results

Study 2 confirmed the results found in Study 1. We found no overall difference in perceived trustworthiness between the instrumental and the categorical value judgement conditions, both for judgements related to higher CO₂ taxes (t_paired(818)= 1.63, p = .10) and for judgements related to higher subsidies for renewable energy (t_paired(818) = -0.75, p = .45). As in Study 1, we found that participants’ prior support for the policy recommended by the scientist strongly predicts perceived trustworthiness, when controlling for the type of value judgement and their interaction for both support for CO₂ taxes ($\text{β}$ = 0.53, SE = 0.03, p < .001) and subsidies ($\text{β}$ = 0.45, SE = 0.04, p < .001). As expected, this relationship is observed for different types of policies (in our case support for taxes and subsidies). Against our hypothesis, but in line with the results from Study 1, we found no significant interaction effect between policy support and condition on perceived trustworthiness, neither for support for CO₂ taxes ($\text{β}$ = 0.01, SE = 0.02, p = .7), nor support for subsidies ($\text{β}$ = −0.01, SE = 0.02, p = .68). Therefore, Study 2 confirms the results from Study 1 that policy support predicts perceived trustworthiness, irrespective of the type of value judgement expressed by the scientist.

When looking at predictors of perceived trustworthiness in the two conditions, we found policy support, climate change concern, and age to predict trustworthiness in both conditions.² As opposed to Study 1, we did not find political orientation to predict trustworthiness perceptions in Study 2 Table 2.

Table 2. Predictors of perceived trustworthiness for the instrumental and the categorical value judgement conditions.

	Perceived trustworthiness
	Instrumental value judgement				Categorical value judgement
Predictors	B	95% CI	$\text{β}$	p	B	95% CI	$\text{β}$	p
(Intercept)	2.00	[1.57, 2.43]		<0.001	1.79	[1.35, 2.24]		<0.001
Policy support	0.26	[0.20, 0.32]	0.31	<0.001	0.25	[0.18, 0.31]	0.28	<0.001
Climate change concern	0.31	[0.24, 0.37]	0.32	<0.001	0.38	[0.31, 0.45]	0.38	<0.001
Male	0.02	[−0.09, 0.13]	0.01	0.748	0.02	[−0.09, 0.14]	0.01	0.671
Political orientation	0.00	[−0.00, 0.00]	0.02	0.494	−0.00	[−0.00, 0.00]	−0.01	0.860
Age	0.01	[0.00, 0.01]	0.12	<0.001	0.01	[0.00, 0.01]	0.11	<0.001
	N = 819, R^2 adjusted = 31.5%				N = 819, R^2 adjusted = 34.8%

Note. Value judgements were averaged across policies (taxes and subsidies). The following measures are presented in this table: Policy support (1 = strongly disagree to 6 = strongly agree); Climate change concern (1 = strongly disagree to 6 = strongly agree); Political orientation (1 = strongly left-wing to 100 = strongly right-wing).

4. Discussion

Scientists have been called upon by policymakers to provide policy recommendations on how to mitigate greenhouse gas emissions and adapt to the consequences of climate change. In their role as policy advisors, scientists often think that they should act according to the value-free ideal of science, and avoid or minimize the use of categorical value judgements. While previous studies investigated only the effect of the communication of instrumental value judgements on climate scientists’ perceived trustworthiness, this study analysed whether the levels of perceived trustworthiness differ between scientists who communicate instrumental and categorical value judgements in a climate change policy recommendation setting. In two studies, we found no differences in perceived trustworthiness between the two conditions but found that perceived trustworthiness perceptions were strongly dependent on whether participants supported the policies recommended by the scientists.

These results have important implications for scientists’ behavior and communication. First, they suggest that climate scientists can recommend policies while challenging existing policy goals (for example, that current climate goals are not sufficient to avoid the worst risks of climate change) without having to fear a loss in perceived trustworthiness compared to simply recommending policies to achieve established policy goals. Second, scientists should be aware that their trustworthiness perceptions are strongly dependent on whether their audiences support the policies they recommend.

We found no significant interaction between our conditions and policy support on overall perceived trustworthiness. It remains unclear if a specific type of value judgement is better suited to preserve trustworthiness when addressing audiences with either low or high policy support. While we observed no differences in overall perceived trustworthiness between our conditions, in Study 1 participants in the categorical value judgement condition perceived the communicating scientist to be less objective than participants in the instrumental value judgement condition. This trend is consistent with previous results (Cologna et al., 2021), which showed that a scientist can be perceived to be less objective, yet still be viewed as trustworthy overall. Authors have argued that treating trust as a unidimensional measure oversimplifies a multidimensional construct and have suggested to use four-factor solutions that specifically measure competence, integrity, benevolence and openness (Besley et al., 2021). Future studies should analyse the effect of different value judgements on individual dimensions of trustworthiness.

When looking at the predictors of perceived trustworthiness, we found that climate change concern and age predicted perceived trustworthiness in both conditions in Study 2, with more concerned and older participants perceiving scientists to be more trustworthy. While we found left-wing political orientation to predict perceived trustworthiness in Study 1, we did not find this effect in Study 2. This could be explained by the fact that our non-representative sample in Study 1 was slightly more left-wing oriented (M = 46.86, SD = 20.31) than our representative sample in Study 2 (M = 48.05, SD = 19.46).

The question remains as to why the use of categorical value judgements did not negatively affect perceived trustworthiness of the communicating scientist. It could be particular to the categorical value judgement chosen in this study (i.e. that current climate goals not sufficient). There is fairly broad public consensus on the necessity to increase action on climate change. In Switzerland, climate change is considered the third most pressing societal issue, with studies showing that 62% of Swiss agree that political action on climate change is necessary (Vimentis, 2019). Public opinion surveys in the US further show that current political efforts to combat climate change are perceived as insufficient (Leiserowitz et al., 2019). Therefore, if scientists express the categorical value judgement that the government should introduce stronger measures than the internationally agreed upon measures, this value judgement is largely in line with the public’s own value judgement. Increasing political action on climate change beyond the internationally established climate goals might thus be seen as a democratic value, which Schroeder (2019, p. 2) defines as “the values held by the public or its representatives.” As argued by Schroeder (2019), when scientists must appeal to values, they should appeal to democratic values. The use of democratic values can increase trust in science as it recovers a different kind of objectivity: “not objectivity as freedom from values, but objectivity as freedom from personal biases” (Schroeder, 2019, p. 21).

Our expressed categorical value judgement does reflect a democratic value: that current internationally agreed upon climate goals are insufficient to address climate change and that national efforts should be scaled up accordingly. Therefore, future studies should try to test the effects of categorical value judgements that are more contested among the majority of the public.

Overall, our results suggest that climate change researchers can make categorical value judgements (i.e. normative assertations with respect to climate policy goals) without having to fear for their perceived trustworthiness, so long as these value judgements are shared with the majority of the public or the audience the scientists are addressing. However, scientists qua policy advisors should be aware that their level of credibility may be contingent on the level of public support for the policy they are recommending or advocating for. Scientists might want to be careful when expressing categorical value judgements, especially concerning policies and policy goals with low public support.

4.1. Limitations and future research suggestions

Some caveats are in order. First, we only surveyed Swiss participants. Past research has shown that public sentiment regarding the role of science in policymaking differs between countries (Cologna et al., 2021). As opposed to Elliott et al. (2017), we did not assess value similarity with the policy goal but rather value similarity with the policy recommendation expressed by the communicating scientist. Future studies should consider this difference when assessing the degree of value similarity between scientists and the public and additionally consider the perceived value similarity in terms of political, religious, or moral values to see how they affect credibility levels.

Given that we did not include a manipulation check, we could not assess whether respondents actually perceived the categorical messages as more normatively laden than the instrumental messages. It could be the case that respondents did not perceive the two value judgements to be mutually exclusive in Study 2; the categorical value judgement could have been perceived to be preceding the need for an instrumental value judgement. Therefore, we cannot exclude the possibility that our manipulation was too subtle, i.e. our tweets were probably perceived as too similar. Further, we specifically tested the communication of different value judgements on the social media platform Twitter by presenting participants tweets. The texts read by participants were therefore quite short; perhaps too short for participants to get a sense of the underlying value judgement. To overcome these limitations, future studies should include manipulation checks and test more than just two short statements. Future studies should also analyse and compare the acceptance of values in different public policy domains (e.g. epidemiology and toxicology) to explore potential differences in more contested domains where democratic values might not be so clear as in the case of climate change. While this study looks at public credibility, future studies could analyse how credibility is affected among policymakers or other groups.

5. Conclusion

We have analysed whether the communication of instrumental and categorical value judgements influences the perceived trustworthiness of a communicating climate scientist. Our data suggests that perceptions of trustworthiness do not suffer when scientists make a categorical, as opposed to an instrumental, value judgement. Given that we found no effect on overall perceived trustworthiness, we believe that climate change researchers can make categorical value judgements without having to fear for their perceived trustworthiness, as long as these value judgements are shared with the audience the scientists are addressing. Support for the policy recommended by the scientist goes along with a high perceived trustworthiness – the specific type of value communication seems to be irrelevant in that case. Our research supports and extends previous findings showing that perceived trustworthiness is dependent upon public support for the policy recommended by a scientist.

Acknowledgements

The authors thank Michael Siegrist (ETH Zurich) for initial comments on the survey used in Study 1 and Niels G. Mede (University of Zurich) for peer-reviewing our analysis code and computing the simulation-based power analysis in Study 2. Viktoria Cologna acknowledges support from the Swiss National Science Foundation Postdoc Mobility Fellowship (P500PS_202935).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are openly available in OSF at https://osf.io/psqea/.

Notes

1 We did not use the item experienced-inexperienced as we told participants that the tweets were shared by renowned scientists, which implies a certain degree of experience. Further, we did not use the item moral-immoral, but responsible-irresponsible instead, as the term “moral” could be negatively associated in the Swiss study context.

2 Note that for these analyses we combined support for both policies into the variable “policy support” as well combining perceived trustworthiness across the instrumental value judgement conditions for both policies and combining perceived trustworthiness across the categorical value judgement conditions for both policies.