There is a growing consensus in philosophy of science that values play an ineliminable role in scientific reasoning. According to Douglas’ (2009) argument from inductive risk, when deciding whether to prioritize avoiding false negatives or false positives in accepting a hypothesis, scientists need to take into the account the social and ethical consequences of these errors. Biddle and Kukla (2017) have argued that epistemic risks exist not just in inductive inferences but also in the construction of scientific models, the operationalization of concepts for scientific research, and data characterization.

The ineliminable role of values in science creates various problems for trust in science. As Wilholt (2013) observed, the cognitive division of labor requires scientists to trust other scientists. However, when each scientist makes these value choices themselves, it becomes difficult for scientists to rely on others and coordinate their research. Wilholt has suggested that methodological conventions help overcome this coordination problem by implicitly imposing a set of value choices on scientists. Irzık and Kurtulmus (2019) have drawn attention to the role values have in bringing about public trust in science. They observed that laypeople will have reasons to consider scientists’ value choices when laypeople need to make practical decisions based on science and when values have a bearing on one’s methodological choices in research. 

While Irzık and Kurtulmus (2019) and Willholt (2013) have discussed the relationship between the role of values in science and trust, they have not offered a normative account of how values should be managed to build public trust in science. This article discusses and evaluates three normative approaches that seek to provide foundations for public trust in science given the ineliminable role of values in science in light of these considerations. 

John’s (2015) high epistemic standards approach maintains that scientists should only share results with the public that are highly certain. This approach enables members of the public to rely on scientists without having to consider their values, since highly certain results would be of use to them whatever their values happen to be. This approach also helps scientists coordinate their research activities and not have to consider the values of the public when conducting their research. Despite its advantages, this approach has two shortcomings. First, it does not help with all kinds of epistemic risks. For instance, it cannot offer guidance on many of the value choices that need to be made in scientific modeling. Second, it can result in the underutilization of scientific findings when high epistemic standards are unattainable despite the science nevertheless being able to provide useful guidance to the public.

According to Schroeder’s (2021) democratic values approach, scientific research should be shaped by the democratic values of a society as determined through democratic processes or surveys. This proposal overcomes both shortcomings of the high epistemic standards approach. It covers all forms of epistemic risk and can also help one make use of science in conditions of uncertainty. However, it has its own problems. The fact that democratic values can be different for each country would make international scientific collaboration more difficult. Given that democratic values can change, Schroeder’s approach also risks creating an intertemporal coordination problem for scientists who need to rely on past research. It also suffers from practical difficulties. Determining the democratic values that should guide scientific research and ensuring that scientists adhere to these values are far from straight forward. The help this approach offers laypeople is also limited. Even if citizens know that research has been conducted in line with democratic values, what this entails for their practical decisions is not easy to tell.

Boulicault and Schroeder (2021) offer a hybrid approach that draws from Wilholt’s (2013) insights into the role of methodological conventions, John’s (2015) high epistemic standards approach, and Schroeder’s (2021) democratic values approach. In cases of low practical importance, the high epistemic standards approach should be used. In situations where scientific knowledge is widely accepted as being highly relevant for practical decisions, the democratic values approach should be used. In cases where research is of moderate practical importance or where disagreement exists about its practical importance, methodological conventions developed by scientist should be used.

This hybrid approach aims to combine the advantages of all three while avoiding their disadvantages. However, given that the value of the high epistemic standards approach relies on its uniform application, it cannot do that. A second difficulty is that research in one category will employ research from others. Thus, keeping the value choices in the different categories isolated as required by this approach would also be impossible.

Although this paper’s primary conclusion is predominantly negative (i.e., none of the examined approaches offer a comprehensive solution), its critical analysis has identified the key criteria a successful solution must fulfill. An effective solution should (a) be applicable by scientists and laypeople who will rely on their findings, (b) enable society to fully utilize available scientific knowledge, (c) address all types of epistemic risks, (d) be compatible with the social structure of science and its cognitive division of labor, and (e) remain incentive-compatible.