Values in environmental research: Citizens’ views of scientists who acknowledge values

Kevin C. Elliott; Aaron M. McCright; Summer Allen; Thomas Dietz

doi:10.1371/journal.pone.0186049

Abstract

Scientists who perform environmental research on policy-relevant topics face challenges when communicating about how values may have influenced their research. This study examines how citizens view scientists who publicly acknowledge values. Specifically, we investigate whether it matters: if citizens share or oppose a scientist’s values, if a scientist’s conclusions seem contrary to or consistent with the scientist’s values, and if a scientist is assessing the state of the science or making a policy recommendation. We conducted two 3x2 factorial design online experiments. Experiment 1 featured a hypothetical scientist assessing the state of the science on the public-health effects of exposure to Bisphenol A (BPA), and Experiment 2 featured a scientist making a policy recommendation on use of BPA. We manipulated whether or not the scientist expressed values and whether the scientist’s conclusion appeared contrary to or consistent with the scientist’s values, and we accounted for whether or not subjects’ values aligned with the scientist’s values. We analyzed our data with ordinary least squares (OLS) regression techniques. Our results provide at least preliminary evidence that acknowledging values may reduce the perceived credibility of scientists within the general public, but this effect differs depending on whether scientists and citizens share values, whether scientists draw conclusions that run contrary to their values, and whether scientists make policy recommendations.

Citation: Elliott KC, McCright AM, Allen S, Dietz T (2017) Values in environmental research: Citizens’ views of scientists who acknowledge values. PLoS ONE 12(10): e0186049. https://doi.org/10.1371/journal.pone.0186049

Editor: Cheryl S. Rosenfeld, University of Missouri Columbia, UNITED STATES

Received: November 27, 2016; Accepted: September 25, 2017; Published: October 25, 2017

Copyright: © 2017 Elliott et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are available on figshare at https://doi.org/10.6084/m9.figshare.4695793.

Funding: This research was funded with support from MSU AgBioResearch to Tom Dietz. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. There was no additional external funding received for this study

Competing interests: The authors have declared that no competing interests exist.

Introduction

Scientists frequently are asked to communicate with policy makers, journalists, and members of the general public on important environmental issues such as climate change, genetically engineered crops, food safety and nutrition, hydraulic fracturing, and pesticide safety. One of the major justifications for federal investment in scientific research is to provide the information our society needs to make well-informed decisions about these issues. Nevertheless, scientists face significant challenges when determining how to communicate effectively and ethically with different audiences. In particular, they must decide how best to maintain their objectivity while providing information that is relevant and accessible to their audiences [1–3].

One of the most significant challenges facing scientists who communicate with the public about major policy-relevant issues is how to address the potential roles that social, cultural, political, and economic values might play in their messages. Scholars distinguish these “non-epistemic” values from “epistemic” values (e.g., predictive accuracy, simplicity, explanatory power) employed within science to evaluate theory and validate knowledge [1]. Scientific objectivity and integrity often have been associated with insulating scientific reasoning from such non-epistemic values [4–7]. Moreover, while many members of the public accord considerable respect to scientists with regard to assessing facts, it is less clear that the public wishes to cede value judgments to scientists [8–10]. However, work in history, philosophy, and sociology of science suggests that non-epistemic value judgments often are intertwined with scientific research in ways that can be difficult to identify and eliminate [1, 11–14]. For example, such values can influence the specific questions asked about a research topic, the types of models and explanations regarded as most compelling and appropriate, the standards of evidence required for drawing conclusions, and the terminology used for describing findings [15–18]. For the remainder of this paper, we use “values” to refer to non-epistemic values.

Given the difficulty of eliminating values from scientific reasoning, some scholars suggest that the best course for scientists is to make their value commitments more transparent [3, 17, 19]. By doing so, they enable their audiences to better understand how such values might be influencing their reasoning, and they alleviate concerns that these values are surreptitiously influencing assessments billed as “value-free.” Recent efforts by many scholarly journals to require financial conflict-of-interest disclosures are one example of how value influences could be made more transparent. In addition, scientists could make deliberate attempts to acknowledge the ethical or social considerations (e.g., economic growth, public health, environmental protection) that they take to be particularly significant when addressing major policy-relevant issues. Even if scientists strive not to let these values influence their reasoning, their audiences still are better equipped to evaluate the scientists’ messages if they understand the values that could be influencing them.

Research objectives

Proposals to promote greater transparency about scientists’ values raise significant empirical questions about how acknowledging values would affect public perceptions of scientists. Some scholars argue that scientists’ credibility could be eroded if they acknowledge their personal value commitments [6, 20]. For example, there is some evidence that when researchers disclose industry funding—which seems to reflect economic-growth values—it decreases perceived research quality and trust [21, 22]. This erosion of credibility may be especially likely among those citizens who do not share the scientists’ specific values or who do not respect the scientists’ research fields. For example, skepticism about scientists’ climate change findings among self-identified conservatives and Republicans in the US general public seems to be fueled at least in part by suspicions that such scientists are influenced by liberal or environmental values they oppose [23–25]. Further, self-identified conservatives report greater trust in scientists conducting research in the service of economic production (i.e., performing “production science”), while liberals place more trust in scientists examining the impacts of that economic production and new technologies (i.e., performing “impact science”) [26].

Other scholars suggest that scientists may increase their credibility by acknowledging their values, because members of the public might respect their efforts to be transparent and honest [17, 27]. This increased credibility may be more likely when citizens share the scientists’ specific values than when they do not [28]. It also may be more likely when scientists provide conclusions that appear divergent from their publicly stated values than when they offer ones that seem expected given their values. For instance, citizens may see scientists who offer conclusions divergent from or contrary to their values as more credible than those scientists who offer conclusions convergent or consistent with their values, because the former seem less likely than the latter to have been guided by motivated cognition [29].

Further, scientists’ credibility might be enhanced or degraded depending on whether the scientists acknowledge values when acting within or outside the conventional purview of science. For example, many hold the normative view that values are irrelevant and even corrupting to scientific reasoning itself, whereas values are necessary when formulating policy recommendations [30–32]. Therefore, citizens may be more willing to accept values statements from scientists acting at or beyond the boundaries of science (i.e., making policy recommendations) than from those acting clearly within those boundaries (i.e., assessing the state of the science) [8, 9].

Our study aims to examine the empirical support for these assertions about how citizens view scientists who publicly acknowledge their values. Briefly, we use the following research questions to frame our study of how a scientist’s public acknowledgement of values influences citizens’ positive affect toward and perceived trust in that scientist. To what extent does it matter:

if the citizens share or oppose the scientist’s values?
if the scientist’s conclusion seems contrary to or consistent with the scientist’s values?
if the scientist is assessing the state of the science or making a policy recommendation?

We conducted two experiments to answer these research questions. Experiment 1 considers a scientist’s interpretation of the state of the science, while Experiment 2 considers a scientist’s policy recommendation. In each experiment, we manipulate whether a scientist’s conclusion appears contrary to or consistent with the scientist’s values, and we account for whether or not subjects’ values align with the scientist’s values. In particular, we instantiate values as broad policy preferences for either promoting economic growth or protecting public health.

Our experiments focus on the health risks of Bisphenol A (BPA), an endocrine-disrupting compound that has come under scrutiny because of its potential to interfere with hormonal activity in the bodies of humans and other species. BPA has been a common additive in plastics, including drinking bottles, medical devices, and the liners of food cans. Recent years have seen considerable scientific controversy about the potential health impacts of its use, with different studies coming to divergent conclusions [33–35]. In 2008, the U.S. Food and Drug Administration declared BPA to be safe, whereas the National Institute of Health’s National Toxicology Program expressed concerns about potential adverse health effects to humans from current exposure levels [36]. The ongoing controversy has been attributed to a number of factors, including differing perspectives on what kinds of biological effects count as “adverse,” efforts by the BPA industry to “manufacture” doubt about its adverse effects, and the hesitancy of regulatory agencies to take action solely on the basis of academic studies [34]. This is precisely the sort of debated, policy-relevant situation in which some scholars have called for scientists to be more transparent about their values [17].

Experiment 1: Interpreting the state of the science

Participants

We administered a survey-based experiment online via SurveyMonkey to participants recruited via Amazon Mechanical Turk (AMT), a crowdsourcing website where “requesters” solicit “workers” to perform “human intelligence tasks” (HITs) for pay. AMT has emerged as a practical way for recruiting a large number of participants from a reasonably wide cross-section of the general public for online experiments [37] or for testing new measurement instruments [38] across the social sciences [39–43].

To solicit a broad cross-section of research participants and minimize self-selection by AMT workers highly interested in BPA or public health, we advertised a HIT titled “Your Attitudes about Important Social Issues in the US.” We limited participation to adults residing in the United States. We paid participants $0.75 for completing the experiment, which took approximately six and a half minutes on average. Compared to a representative sample of the US general public, our AMT convenience sample is more male, younger, more highly educated, and more liberal. The sample contains 494 subjects who completed the entire experiment (of the 500 who began it) on September 17, 2015.

The experiment

Our experiment employed a 3x2 full factorial design: values (values not mentioned, economic growth values present, or public health values present) by conclusion (BPA causes harm, BPA doesn’t cause harm). This was a fully randomized design that did not take into account any characteristics of the subjects. While in future research some stratification in the randomization process might be useful (e.g., to balance sample size across some of our key non-experimental independent variables), we had no a priori information that would guide a more complex design of this sort. The exact stimulus messages presented in each of the six resulting conditions are presented in Part A of S1 File. In our analyses we divided subjects into those who favored promoting economic growth over protecting public health and those who favored protecting public health over promoting economic growth and then examined the effects of our experimental manipulations within those groups.

The Social Science Behavioral/Education Institutional Review Board (SIRB) of the Human Research Protection Program at Michigan State University approved this research (IRB#x13-556e) (Exempt 2). After reading a consent page and agreeing to participate in our study, subjects answered questions about their preference for promoting economic growth or protecting public health and about whether they view certain behaviors by scientists as ethical or unethical. They then were randomly assigned to one of the six conditions above. All subjects received the following background information before reading their brief experimental message:

For several decades, a scientist named Dr. Riley Spence has been doing research on chemicals used in consumer products. One chemical—Bisphenol A, popularly known as BPA—is found in a large number of consumer products and is suspected of posing risks to human health. Yet, scientists do not agree about these possible health risks. Dr. Spence recently gave a public talk in Washington, D.C. about BPA research. Here is the final slide from his presentation.

After reading their assigned stimulus message, subjects answered two open-ended comprehension questions and two closed-ended manipulation check questions. Subjects then answered two sets of questions about their perceptions of Dr. Spence. After subjects answered a few brief demographic, social, and political questions, we thanked them for their participation and debriefed them about our research question.

Variables

We employed two composite outcome variables in our analyses. Positive affect (Cronbach’s alpha = 0.96) is a semantic differential scale that is the average of responses to seven items, each of which was measured on a seven-point scale. It measures the positive affect that subjects have toward Dr. Spence. The seven items, whose response categories ranged from “not at all” = 1 to “very” = 7 were: competent, credible, expert, honest, intelligent, sincere, and trustworthy. Perceived trust (Cronbach’s alpha = 0.98) is a scale that is the average of seven measures of perceived trust, each of which was measured on a seven-point scale (“completely distrust” = 1 to “completely trust” = 7) using the following items: draw scientific conclusions that are independent from his research funders; draw scientific conclusions that are independent from his own political values; fully consider other scientific studies with results that don’t match his political values; fairly report all sides of BPA research to the news media; fairly report all sides of BPA research to Congress; clearly represent uncertainties about the effects of BPA to the news media; and clearly represent uncertainties about the effects of BPA to Congress.

We used three dummy variables to account for our experimental conditions. Two dummy variables distinguish whether the scientist was labeled as favoring economic growth (scientist values economic growth) or as favoring protecting public health (scientist values public health) with “no values present” as the reference category. A third dummy variable differentiates the condition where the scientist concludes that BPA causes harm (scientist concludes BPA causes harm) from the one where the “scientist concludes BPA does not cause harm,” which serves as the reference category.

We assessed subjects’ pre-existing preference for either promoting economic growth or protecting public health with the following question: “National policies are often the result of trade-offs between important priorities. When there is a tradeoff between promoting economic growth and protecting public health, which BEST captures your preference?” The four response categories (“strong preference for promoting economic growth,” “mild preference for promoting economic growth,” “mild preference for protecting public health,” and “strong preference for protecting public health”) were dichotomized to represent a preference for either promoting economic growth (37.1% of subjects) or protecting public health (62.9% of subjects).

We also employed four demographic, social, and political variables as statistical controls in our analyses. Female is a dummy variable for gender: “male” = 0 and “female” = 1. Approximately 52% of our subjects are male. Age is measured with ten categories: “18–19” = 1 to “60 or higher” = 10. Approximately 75% of our subjects are younger than 40 years old. Education is measured by the highest degree earned: “up to high school diploma or equivalent” = 1 to “graduate/professional degree” = 4. Approximately 53% of our subjects have a bachelor’s degree. Political ideology is measured on a 7-point scale from “very conservative” = 1 to “very liberal” = 7, with “middle-of-the-road” = 4. Approximately 52% of our subjects identify as slightly to very liberal. We treat age, education and political ideology as continuous variables to preserve degrees of freedom. Thus, our analysis may not capture any strongly non-linear effects of these variables.

Analytical techniques

We analyzed our data with a series of ordinary least squares (OLS) regression models using the Stata 14.1 software package. We also estimated Huber-White standard errors that are robust with regard to non-normality and heteroscedasticity; those estimates yielded substantively identical results and are not reported here. Since the positive affect scale and the perceived trust scale each take on a large number of distinct values (64 and 43, respectively), we felt justified in treating these as continuous outcome variables in OLS regression. All of these models included our four demographic, social, and political variables as controls. For each outcome variable (positive affect and perceived trust), we ran a model examining the main effects of the following four predictor variables: scientist values economic growth and scientist values public health with “no values present” as the reference category; scientist concludes BPA causes harm with “scientist concludes BPA does not cause harm” as the reference category; and subject favors public health with “subject favors economic growth” as the reference category.

To examine those interaction effects suggested in our research questions, we ran OLS regression models predicting each outcome variable with the two values statements dummy variables (scientist values economic growth and scientist values public health) within four subsamples. We defined these four subsamples by the scientist’s conclusion about the state of the science on BPA (whether the scientist concludes BPA causes harm or concludes BPA does not cause harm) and the subject’s own value preference (whether the subject favors promoting economic growth versus favors protecting public health). Running separate analyses within subgroups allows each subgroup to have its own estimated error variance but reduces degrees of freedom for each analysis. An alternative suggested by a reviewer is to use all cases to preserve degrees of freedom, and thus power, using interaction effects to allow coefficients to vary across subgroups. However, this approach constrains estimated error variances to be equal across subgroups and introduces high collinearity among the interaction terms. We also have conducted the analysis using interaction effects and all cases, and the results are essentially identical to those reported here.

We report p values as is standard practice. In addition to the standard cautions on the use of p values [44], we emphasize the following. Since we are using a convenience sample, while p has the usual interpretation for the experimentally manipulated variables, the p value for other variables is best interpreted as a comparison to a purely random variable. As suggested by a reviewer, we also note that the power of statistical tests in the subgroups is less than ideal. We emphasize again that our results should be viewed as exploratory and in need of replication.

Results

Tables 1 and 2 present the results of OLS regression models explaining positive affect toward and perceived trust in the scientist who offered an interpretation of the state of the science on BPA, respectively. In each table, the first model estimates main effects in the full sample and the four subsequent models estimate the effects of the two values dummy variables in each of four subsamples. Figs 1 and 2 visually represent selected results from the tables.

Download:

Fig 1. Effects of scientists’ conclusions and expressed values on respondents’ positive affect toward them.

Effects are relative to a scientist not expressing a preference for particular values and are net of control variables; see Table 1. Green arrows are used when the scientist’s conclusion coincides with the scientist’s values; red arrows are used when the scientist’s conclusion conflicts with the scientist’s values.

https://doi.org/10.1371/journal.pone.0186049.g001

Download:

Fig 2. Effects of scientists’ conclusions and expressed values on respondents’ trust in them.

Effects are relative to a scientist not expressing a preference for particular values and are net of control variables; see Table 1. Green arrows are used when the scientist’s conclusion coincides with the scientist’s values; red arrows are used when the scientist’s conclusion conflicts with the scientist’s values.

https://doi.org/10.1371/journal.pone.0186049.g002

Download:

Table 1. Measures of the effect of scientists’ conclusions and expressed values on respondents’ positive attitude toward them (Experiment 1).

https://doi.org/10.1371/journal.pone.0186049.t001

Download:

Table 2. Measures of the effect of scientists’ conclusions and expressed values on respondents’ trust (Experiment 1).

https://doi.org/10.1371/journal.pone.0186049.t002

In the two overall regression models with only additive effects, subjects exposed to a scientist mentioning either economic growth values or public health values report lesser positive affect toward and perceived trust in the scientist than do subjects exposed to a scientist mentioning no values. The effect of the presence of economic growth values is substantially greater than the effect of the presence of public health values both on positive affect (F_1,483 = 9.01, p = 0.003) and on perceived trust (F_1,483 = 5.95, p = 0.015). Also, subjects exposed to a scientist who concludes that BPA causes harm report greater positive affect toward and perceived trust in the scientist than do subjects exposed to a scientist who concludes that BPA does not cause harm. Subjects’ own values (i.e., protecting public health versus promoting economic growth) have no effect on either positive affect or perceived trust. Also, age is the only social, demographic, or political control variable that has a statistically significant effect on either outcome variable in the overall sample, with older subjects reporting greater positive affect toward the scientist than their younger counterparts. At the suggestion of a reviewer, we also re-estimated this model without controlling for political ideology and found no substantial changes in our results.

Shifting to the results of the models with the four subsamples allows us to explore the more nuanced patterns suggested by the scholars’ views mentioned earlier. Exposure to a scientist acknowledging economic growth values significantly decreases positive affect toward and perceived trust in that scientist for subjects in three of four subsamples. The exception is the group of subjects who favor economic growth over public health and who were exposed to a scientist concluding that BPA causes harm. Exposure to a scientist acknowledging public health values significantly decreases positive affect toward and perceived trust in that scientist only for subjects who favor public health over economic growth and who were exposed to a scientist who concludes that BPA causes harm.