Trust, trustworthiness, and relationships: ontological reflections on public trust in science

ABSTRACT

There is much social scientific research dedicated to measuring and studying public trust. I examine the ways in which the notion of trust is implicitly conceptualised in such studies. I argue that there is a common ontological foundation in most research on the topic: trust is viewed as a phenomenon that is an attribute of individuals, intrapsychic, directed toward specific targets, and that is quantitative and measurable. I criticise this conceptualisation of trust and argue that it: (1) fails to consider the trustworthiness of individuals and institutions, (2) fails to recognise trust as a relational phenomenon and overlooks historical and material conditions that characterise relationships between people and institutions, and (3) lends itself to bureaucratic manipulation of publics rather than fostering authentic relationships of trust. I conclude that studies on trust need to be situated in larger frameworks that attend to the trustworthiness of actors and to relationships between them.

KEYWORDS:

• Public understanding of science
• ontology
• trustworthiness
• deficit model
• democracy
• psychological measurement

Trust, distrust, and democracy

Trust is essential to collective self-rule. In representational democracies leaders do not hold their positions based on divine right, rule of force, or familial ties; rather, they are elected for limited terms by their constituents to represent and act on their interests. The legitimacy of the positions of leaders is thus supported by the trust of their constituents that elected leaders will, indeed, act on behalf of their collective interests and benefit. When this is not the case, for instance when religious or ethnic minorities feel that political leaders do not represent their interests or when elected officials fail to act on their campaign promises, the resulting distrust is evident. More broadly, distrust in politicians or distrust in democratic systems is associated with political apathy and protest (Hooghe and Marien 2013). Warren (2018) argues similarly that trust is an integral aspect of democratic governance since trust relationships are the building blocks of social cooperation. There are simply too many decisions that need to be made at a governmental level for all citizens to attend to or even be aware of. There is thus benefit in sharing participatory labour: a small number of individuals can dedicate themselves to one or a few political issues, while trusting others to dedicate themselves to other issues. Such focusing of attention could be formal, as in the case of elected officials or professional civil servants, or more informal as in the case of deliberative approaches to democracy (see O’Doherty 2017). In spite of the reliance of democratic systems on trust, they are also implicitly premised on distrust (Tilly 2007; Warren 2018). A key function of democratic government is to make decisions in cases of conflict (over resources, beliefs, values). Not only is distrust a likely characteristic of relationships between stakeholders on opposing sides of such conflicts, the governmental role in making decisions in such cases is itself not necessarily conducive to building trust. Warren (2018) thus argues that democratic institutions structure both trust and mistrust by (1) protecting generalised trust (for example, through recognising and enforcing the protection of rights), (2) separating political institutions (e.g. political parties) from institutions that hold public trust (e.g. agencies that provide security and welfare), (3) channelling distrust by, for example, recognising conflicting interests of different groups and possibilities for resolution, and (4) by empowering citizens to monitor governmental actors. In short, trust is a foundational aspect of living in a democratic society and democratic systems serve a critical role in structuring distrust.

Moral philosophers have similarly argued for the importance of trust in governing human relations. Baier’s (1986) analysis of trust is particularly pertinent as it engages with questions of when trust is warranted and when it is not. Her analysis highlights the importance of considering the contexts within which trust relationships exist, and the power asymmetries inherent in such relationships. Similarly, McLeod (2020) discusses the epistemology of trust as involving questions of ‘Ought I to trust?’ and ‘Ought I to distrust?,’ though she observes that the latter question has received insufficient attention among philosophers. Compared to the more descriptive accounts of trust that tend to be favoured by psychologists and political scientists, philosophical scholarship has thus tended to focus on when we should trust or distrust (Faulkner and Simpson 2017).

Science, technology, and trust

Science and technology are human and social activities that are distributed and hold different meanings in different contexts. As such, to speak of ‘trust in science’ is not very meaningful (Resnik 2011). Trust in whose science? Trust in which proclamations that have been labelled ‘scientific’? Trust in which scientific institutions? Science is not a single object; it is not even a specific process or method that can easily be pointed to (Ankeny n.d.; Jasanoff 2004). The term ‘science’ itself is polysemous, and there is great discursive diversity in how the term can be deployed by even the same speaker in different conversational contexts (Gilbert and Mulkay 1984). In Science and Technology Studies (STS) and related fields these insights are considered truisms: science and technology are produced by humans, shaped by social forces, in particular cultural contexts, with their own histories and in particular historical periods. However, outside of STS and cognate fields, and in most non-academic settings, science is often considered to be very simply a privileged mode of truth telling, and technologies are considered to be objects that improve life by enabling new actions or solving problems in a somewhat linear fashion. Thus, when we talk about public responses to science and technology, we tend to think of ‘science and technology’ in this latter way, namely as a symbolic category consisting of imagined agents such as ‘scientists,’ ‘technology companies,’ ‘genetics,’ and ‘computers,’ as well as the attributions that might be associated with these terms.

Both trust in science, in general, and trust in particular areas of science and technology have been topics occupying scholars and governments. A key reason for this is an ongoing perception of a decrease in public trust in science (Wynne 2006). Similar to democratic governance, scientific practices may be seen to institutionalise mistrust, for example in the form of academic review. However, when it comes to the (lack of) public trust in science and technology, commonly voiced concerns tend to presume that science is something that should be trusted, and that deviations from this default trust are something to be explained and rectified (Contessa 2022, calls this the naïve account of public trust in science). Because of the privileged role that science has been accorded as a source for truth telling, the default position presumed by scientists and government agencies heavily reliant on science for the justification of policies (e.g. public health agencies) is that the pronouncements of ‘science’ deserve to be, and therefore, should be trusted. I will return to this problem below.

What is trust?

I now turn to the main questions that motivate this article, namely what is trust and how is trust conceived in empirical studies? Given the importance of the role that trust has not only in the application and communication of science, but in almost all social interactions, it is not surprising that there is a very large amount of empirical research on trust.

Definitions have been developed by scholars for not only trust, but for several distinct kinds of trust (there are too many to list but see, for example, Giddens 1990; McLeod 2020; Seligman 1997). However, definitions do not solve the problem. For any one definition developed by one scholar, another scholar can develop a slightly different or even opposing definition. So, in the face of competing definitions what, really, is trust? Competing definitions cannot be evaluated for their proximity to ‘the truth’ because there is no concrete object commonly known as ‘trust’ against which definitions can be compared. The best that could be done is to compare operational and scientific definitions of trust with common and everyday uses of the term. Just like the term ‘science,’ the term ‘trust’ is highly polysemic (Jesuino 2008). In some instances, the term is used synonymously with other terms such as confidence, reliance, expectation, or solidarity. In other instances, it has different connotations from those terms (Markova, Linell, and Gillespie 2008). No matter how sophisticated the development of definitions and operationalisations of trust is, the problem remains that trust has no tangible referent (i.e. there is no object that is unambiguously associated with the term ‘trust’). For this reason, without deeper conceptual analysis, even scientific treatments of trust tend to rely on the assumption of an ontology underlying particular uses of the term ‘trust’ that are at best inferred, but more often simply taken for granted.

Why is this problematic? I hope to show in this article that the way in which we conceive of trust has important ramifications for how we study trust empirically and, thus, for the knowledge claims that are made about trust in the context of science and society. Such knowledge claims about public trust in science are produced by researchers with applied intentions and so they have relevance for policy, for how particular publics are portrayed and treated by governments, and for the ways in which relationships between/among publics and science evolve.

The purpose of this article is thus to engage with the ontological foundations of notions of trust.¹ A major challenge in doing so is that there is so much scholarship on trust, both empirical and theoretical, and in several disciplines, that it is not possible to provide a review that does justice to the topic. In the first instance, therefore, my focus is on empirical research on the topic of trust in science and technology, with a particular focus on recent research conducted in the context of COVID-19. I do contend, however, that the criticisms I develop in this context are generalisable to a wider range of contexts in which trust is studied empirically, and even some philosophical approaches to trust. As such, I argue that the problems in research on trust in the context of COVID-19 are illustrative of a much bigger problem. Second, there is a great deal of attention that has been given to understanding trust from a theoretical perspective primarily in philosophy. I, therefore, draw on relevant theorists from philosophy as I develop my argument, without claiming to be comprehensive in my treatment.

My main argument is that prevalent approaches to trust view it (often implicitly) as an intra-psychic phenomenon. Such studies thereby neglect important normative considerations of whether particular actors should be trusted. They also neglect examination of the relationships between actors which is problematic especially in situations where such relationships are characterised by unequal power and by past harms (see also Baier 1986). I argue that we need to develop more comprehensive and pluralistic ways of understanding and studying trust. In particular, ‘trust’ should not be thought of as an object, but rather as a phenomenon with at least three distinct aspects: a psychological aspect (the trust felt be a person experientially), a normative aspect (whether someone should be trusted or not), and a relational aspect (the nature of the relationship between a ‘truster’ and the party they do or do not trust). Empirical studies of trust in science tend to ignore the question of trustworthiness of science and the history of the relationship between different groups and particular scientific actors or institutions. In doing so, these studies often end up adopting an approach that lends itself to the manipulation of publics.

The structure of my argument is as follows:

1. In spite of wide variation in how trust is defined in empirical studies, there is a dominant ontological foundation to how trust is conceived that is widely prevalent.

2. This common ontological foundation underlying many studies holds that trust is:

   • The property of individuals.

   • Intra-psychic (i.e. located in people’s minds).

   • Directed at specific targets (e.g. how much do I trust this particular person/institution).

   • A phenomenon with a quantitative structure.

3. I then criticise this ontological foundation of trust because it:

   • Fails to give attention to trustworthiness.

   • Fails to recognise the relational nature of trust.

   • Lends itself to manipulation, rather than meaningful dialogue.

I conclude by arguing that trust is better understood not as a single object or construct, but rather as a broader phenomenon with several aspects. Trust, therefore, needs to be studied in a pluralistic way to enable a more comprehensive understanding.

How do researchers conceptualise trust: the case of COVID-19

Given the pre-occupation with public trust by government actors and scientific institutions, there are many topics that have been the focus of research on trust. Notable among these are controversial issues such as GM foods (Augoustinos, Crabb, and Shepherd 2009), nuclear energy (Bratt 2015), the BSE crisis in the UK (Jones 2004), and human genetics (Jones and Salter 2003). For this part of the discussion, I will focus on research on trust as it relates to COVID-19. Patterns of how researchers have approached studies of trust are surprisingly similar across topics. Nevertheless, COVID-19 is a particularly interesting case to examine because of the global reach of the pandemic, the nature of governmental responses to curb it, and notable commonalities and differences in responses across jurisdictions. As might be expected, public responses to governmental actions have been just as (or even more) complex and often polarised. Almost all governmental responses from the beginning of the pandemic (early 2020) to date (August 2021) have been intertwined with controversies relating to scientific expertise and the use of technologies to combat the virus and its spread. Here are just some examples:

• Governments’ reliance on ‘scientific advice’ to argue against the need for wearing masks early in the pandemic (at least in places like Canada), followed by strong mandates for mask wearing in many jurisdictions based on updated scientific advice

• Mandated social distancing measures to curb transmission of the virus, based on research on the transmissibility of the virus

• Various stages and degrees of lockdowns, requiring people to stay in their homes, often based on mathematical models and arguments for ‘flattening the curve’ and to avoid overwhelming hospitals and health care systems

• The use of digital technologies for contact tracing

• The use of genome sequencing technologies for tracking mutations of the virus

• The rapid development and deployment of novel types of vaccinations to protect against COVID-19 symptoms

The rhetorical use of scientific expertise in justifying specific actions by governments and public health agencies has been particularly interesting to observe in cases where different jurisdictions have implemented drastically different responses (for example, relating to school closures: https://ourworldindata.org/policy-responses-covid). Presumably, ‘the science’ is not all that different across different political jurisdictions, and yet ‘science’ and ‘listening to the scientists’ is often a primary justification to support very different actions by different governments. In the UK, in particular, rhetorical use of ‘science’ to support actions taken by the government became a point of controversy, raising difficult questions about who has the authority to speak on behalf of ‘science’ (Thacker 2020). The introduction of new technologies to combat the virus and curb its spread, such as mRNA vaccines and digital contact tracing apps, have also been associated with much controversy among publics and experts (Nelson 2021; Lucivero, Hallowell, and Johnson 2020). This is not to say that governments and public health agencies have inevitably suffered from a loss of public trust as a result of COVID-19 and their efforts to address its societal effects. In some instances, governments’ responses to the pandemic have succeeded in achieving quite staggering new heights of approval from their publics, as has been the case for New Zealand’s prime minster, Jacinda Ardern (McClure 2021). The COVID-19 pandemic has required governments to respond to circumstances with measures that have had ramifications on trust relationships between individual citizens, groups, and institutions. Unsurprisingly, this situation has been accompanied by numerous scholarly studies on trust and COVID-19.

We conducted a review of peer-reviewed literature of all studies that dealt with trust in the context of COVID-19 from the start of the pandemic to April 2021. We found 90 articles that examined some aspect of the role of trust in public responses to COVID-19 and measures taken to curb the pandemic. Across these articles, definitions of trust and implicit assumptions about what kind of ‘thing’ trust is, varied quite dramatically. Many studies explicitly measured trust but did not refer to or provide a definition of the particular conceptualisation of trust being used. For example, Sherman et al. (2020) write simply that in their study ‘[s]tatements measured theoretical constructs including … trust in the Government and in the NHS’ (1613). Similarly, Oksanen et al. (2020) write about the importance of ‘institutional trust’ in pandemic management. In these cases, researchers provided no explanation of what kind of construct of trust they were employing, though it was generally possible to infer at least some aspects of the ontology associated with ‘trust.’

In many studies, researchers did provide definitions of trust, though conceptualisations of trust used across studies were often at odds with each other. In particular, some researchers associated trust with personality or dispositions of individuals and thus as a relatively stable quantity within individuals. For example, Ye and Lyu (2020) view ‘generalized trust’ as an ‘optimistic trait’ (2). Alessandri et al. (2020) provide a more detailed review of a body of work that views trust as dispositional and as ‘a downstream consequence of proximate dispositions such as personality’ (4) but, following Uslaner, 2012, define ‘generalized social trust’ as a belief (‘the belief that most people can be trusted’ – 2). In contrast to the few studies that implied trust to be a somewhat stable attribute of individuals (disposition or trait), more generally researchers tended to view trust as a state or variable quantity. For example, in a study that sought to understand the relationship between vaccine hesitancy and trust, Palamenghi et al. (2020) measured the change of individuals’ ‘trust towards science.’ Roozenbeek et al. (2020) similarly treated trust as a variable in survey items such as trust in the government, scientists, and journalists.

Researchers also varied as to whether trust was conceived of as a consciously accessible psychological state or belief, or an unconscious attribute of an individual. By far the most common approach was for researchers to treat trust as a psychological attribute that could be measured simply by asking individuals about their degree of trust in a particular person, institution, or class of people. Trust was thus treated as an intrapsychic construct that is consciously accessible to individuals for assessment and that individuals can accurately report on their own trust levels. This implicit understanding of trust is matched by its measurement using surveys and Likert scales in most studies. Even in studies that investigated the operation of non-conscious psychological mechanisms, such as cognitive styles, ‘trust’ was treated as consciously accessible. For example, Fuhrer and Cova (2020) examined cognitive style (analytic versus intuitive) and trust in Didier Raoult, a prominent French physician. The latter was measured using a series of questions such as ‘How much do you trust Professor Raoult? (−3 = Not at all, 3 = Completely).’ An interesting exception to the trend of studying trust via survey questions was a study by Li, Zhang, and Niu (2021) who compared pre- and post-pandemic trust behaviour. This was one of the few studies that studied trust in an experimental framework that measured trust as a behaviour in an online trust game, rather than a reported psychological state or belief.

Finally, researchers posited the existence of different kinds of trust that they then investigated. These included information trust (Laato et al. 2020), social trust (Min 2020), conservative trust and liberal trust (Agley 2020), source trust (Lu et al. 2021), and vaccine trust (Latkin et al. 2021).

Common ontological foundation of trust

In spite of the evident variation in conceptions of trust across studies, there was a notable common ontological foundation. Research that incorporated measures of trust, with very few exceptions, was based on a set of assumptions about what kind of ‘object’ trust is:

• Trust was treated as an individualistic construct, in the sense that it was seen to be a property of individuals.

• Studies used a psychological conception of trust; more specifically, trust was presumed to be an intrapsychic phenomenon.

• Trust was seen to be directed toward particular referents (government, scientists, public institutions, health services, politicians, journalists, etc.).

• The phenomenon of trust was implicitly presumed to have a quantitative structure, such that individuals can have more or less of it directed towards a specific referent.

I now provide more detail on each of these points.

Trust as individualistic construct

A particularly strong theme in researchers’ conceptualisation and treatment of trust is that it is the property of an individual, irrespective of whether stable or variable. Even in cases where researchers drew on a theoretical foundation for trust that implied a collective or social ontology of the construct, trust was treated linguistically and operationally as an individual construct (see also Contessa 2022, on individual approaches to trust). For example, Min (2020) reported a study about the role of social trust in slowing down or speeding up the transmission of COVID-19. The study’s premise and conclusions are framed in terms of countries as a unit of analysis. For instance, ‘in countries characterized by high levels of social trust (especially the particular social trust) or a narrower or wider range of trustees, the number of new cases is likely to reach the peak faster than in other countries’ (14). Nevertheless, the definitions of trust reviewed by the author emphasise trust as an attribute of individuals (not societies) (e.g. social trust ‘has been defined as the belief that other people will not cause any harm to us but will rather look after our interests’ and it is a ‘psychological state’ and an ‘expectancy’ – 2). Accordingly, the study used measures of trust based on individuals’ responses to questions about how much they trust particular entities (e.g. ‘How much do you trust people from various group? Your family; your neighbourhood; people you know personally; people you meet for the first time; people of another religion; people of another nationality’). In short, trust was overwhelmingly treated as a phenomenon associated with individuals. Even in those cases where researchers spoke of trust as being associated with countries or groups, the specific definitions and measures of trust that were used implicitly positioned trust as an attribute of individuals.

Trust as intrapsychic

Second, trust was implicitly treated as an intrapsychic phenomenon. The fact that many might respond to this statement by asking, what else would it be if not an intrapsychic phenomenon, is perhaps telling as to the degree to which studies of trust rely on taken-for-granted assumptions, rather than careful conceptual work. The obvious alternative is that the ontological foundation of trust could lie in the relationship between people, that it is intersubjective rather than intrapsychic. More about that later. At this point, I simply want to point out that in spite of some disagreement about whether trust is more about beliefs or emotions, almost all empirical research on trust is based on an understanding of trust as located in psychic aspects of persons. This is most evident in studies that used surveys or scales to ask individuals about the level of trust they hold toward targets, but is also implicit in studies that used other approaches to measure levels of trust.

Trust is directed to specific referents

The third common element to the dominant conception of trust is that trust is directed toward specific referents. That is, I may have high trust in scientists and low trust in politicians, or high trust in qualitative researchers and low trust in researchers who use statistics. Certainly, some studies also conceived of trust having dispositional aspects that were independent of specific targets. For example, researchers who view trust as a personality trait would conceive of some people as inherently more trusting than others. However, even in such cases this was generally seen as an underlying predisposition, and individuals would still be seen as having higher or lower levels of trust toward particular targets, irrespective of whether they had ‘trusting’ personalities. One exception is the notion of generalised trust, which was often used to denote an intrapsychic attribute not directed at a specific target.

Trust as quantitative and measurable psychological phenomenon

Finally, trust was conceived as a phenomenon that is quantifiable, as something that we have as individuals in greater or lesser quantities toward a particular target. What that ‘something’ is of which we have more or less, is never quite clear. For instance, the metaphor of a water glass which has a certain amount of liquid in it does not quite seem to fit the conceptualisation of trust implied by trust researchers; that is, trust does not seem to be thought of as a substance of which there is more or less. This begs the question, if trust is quantitative what, specifically, is there more of or less of?

The problem of measurement in psychology (and of psychological attributes) has received much attention over the decades. Strong criticisms of empirical researchers’ assumptions about the (putative) quantitative structure of psychological attributes have been developed by several theorists (e.g. Michell 2000; Tafreshi, Slaney, and Neufeld 2016; Trendler 2013; Franz 2022). These theorists point out that there is a fundamental problem in assuming a priori that a particular psychological phenomenon has a quantitative structure. They disagree about the precise ways in which one might conclude that a particular psychological attribute does, indeed, have a quantitative structure. Michell (2000) argues that it is necessary to establish empirically that a phenomenon has a quantitative structure; Franz (2022) argues that this needs to be determined by conceptual analysis; Trendler (2013) argues that it is not possible in practice to determine whether psychological attributes have a quantitative structure. What these theorists do agree on, is their observation that many researchers fail to pay attention to the question of whether there are grounds to support the supposition that phenomena associated with specific psychological constructs being studied have a quantitative structure, and yet nevertheless go about with much enthusiasm ‘measuring’ levels of, say, trust, using instruments such as surveys and Likert scales.² To my knowledge, no conceptual or empirical study has been conducted to determine the existence of a natural phenomenon associated with the term ‘trust,’ located inside individuals, that has a quantitative structure.

To sum up, a common ontological foundation of trust posits that trust is an intrapsychic phenomenon, that is quantifiable, and that is located within individuals and specific to particular targets. There are some exceptions to this dominant conception, and I discuss these below.

Criticism of individualistic and intra-psychic conceptualisations of trust

Having characterised what seems to be a common and dominant conceptualisation of trust, I now develop several criticisms of this approach. I argue that, in addition to the epistemological and measurement problems I have outlined above, there are three kinds of problems with studies that are based on this conceptualisation of trust:

• Failing to examine the trustworthiness of individuals and institutions

• Failing to recognise the relational nature of trust

• Purpose of research to manipulate people

In developing my criticisms, I am mindful not to set up straw man arguments. It is thus worth emphasising that I am drawing attention to a dominant trend in empirical research on trust, not making universalistic claims about all such studies. Further, some of my criticisms rely on implicit treatments of trust; that is, in the absence of explicit definitions of trust, my analysis is based on what kind of measures of trust are used, how trust is written about, and how trust is built into empirical studies. I have no direct access to researchers’ beliefs or assumptions about trust; my commentary and criticism is confined to how it is written about in published work. In recognising that I am only pointing to a general trend, I also note several studies explicitly that are exceptions to the common ontological foundation I outlined above to illustrate possible alternative ways of studying trust.

Trust needs trustworthiness

When the focus is primarily or exclusively on the ‘trust’ that an individual holds, there is the risk of failing to consider whether the corresponding target is worthy of being trusted. When this occurs, the focus of measurement on trust as a psychological phenomenon narrows attention on the individual doing (or failing to do) the trusting, without considering whether they should be trusting. The connection between trust and trustworthiness has been famously studied by Onora O’Neill (2017) who uses the compelling example of Bernie Madoff to illustrate her point. Madoff was an American financier who ran an investment fraud that cost victims billions of dollars. Madoff’s clients clearly trusted him with their money, and it is uncontroversial to claim that in this instance their trust was misplaced; Madoff was not deserving of their trust. Of course, the issues studied by researchers who measure trust are somewhat different. In public health and related domains of policy, the targets of trust that are of interest to researchers tend to be scientists, doctors, governments, health institutions and agencies, and similar entities. That is, they tend to be actors in positions of authority who one wants to presume are trustworthy. However, arguably, the same might have been said about Madoff who had held positions of authority such as the chairman of the NASDAQ stock exchange. Scientists are not always right (note, for example, growing list of cases of academic fraud), doctors make mistakes (see, for example, literature on medical error), politicians and governments don’t always act in the interests of their constituents (no examples needed), and health systems often fail to provide equitable care to individuals (see, for example, issues with systemic racism in healthcare systems of many countries). In other words, whether any given institution or person is deserving of trust in a particular case and time, is not something that can or should be assumed. As O’Neill (2002) explains, a key challenge relating to trust is to know when to be trusting and when to be mistrustful.

The problem of trustworthiness in the context of science and society is well recognised by scholars in the philosophy of science. Whyte and Crease (2010), for example, argue for the importance of normative conceptions of trust in examining relations between science and society. In that vein, John (2017) examines under what conditions publics should trust the scientific claims of experts and (2018) questions whether the norms of transparency, openness, sincerity and honesty should be seen to contribute to scientists’ trustworthiness, coming to the conclusion that they do not. Similarly, Irzik and Kurtulmus (2019) examine the conditions under which trust in scientists is warranted, and Almassi (2012) argues for reciprocal affective trustworthiness between scientists and lay publics. My point here is not to engage in discussion about how specifically the trustworthiness of scientists, scientific institutions, or scientific claims should be assessed, but rather to point out that there is broad agreement among philosophers of science that it is crucial to give attention to trustworthiness of science.

In spite of this, Wynne (2006) has observed that many responses of institutional science to perceived problems of ‘public mistrust in science’ tend to attribute blame to ignorant publics and misinforming media, rather than reflecting on how institutional science might itself be implicated in leading to conditions that foster distrust (212). Research that is focused on trust as a psychological attribute tends to leave the trustworthiness of actors in particular contexts unexamined. In measuring only how much trust individuals hold toward specific actors, such studies fail to consider how much trust one should have toward particular actors in particular instances.

A focus on trustworthiness as opposed to trust has important implications for both the person doing the trusting and the entity asking to be trusted. For the person doing (or not doing) the trusting, focusing on trustworthiness allows for careful discrimination in deciding what situations and relationship require trust versus mistrust. For individuals or institutions asking to be trusted, the focus on trustworthiness leads to an inward shift in attention, which allows for reflection on one’s own (individual or institutional) conduct and how this is implicated in fostering trust or mistrust. For example, in the context of public trust and vaccination, Goldenberg (2021) argues that trustworthiness can be fostered not by shouting down mavericks who speak out against vaccination, but rather by using the ascendancy of maverick views as an opportunity for experts and expert institutions to reflect on their own shortcomings and through addressing those earn public trust (167).

It is important to note that the term ‘trustworthiness’ is sometimes used in a different sense from what I am referring to here. For example, in some psychological research ‘trustworthiness’ is used to refer to traits or characteristics of a person or institution that will lead people to trust them. Trustworthiness in these kinds of studies is thus seen to be an attribute or a perceived attribute of individuals. Similarly, several studies that we examined in the context of COVID-19 and trust did refer to trustworthiness. For example, Lu et al. (2021) examined perceived trustworthiness and Okan et al. (2020) and Malesza and Kaczmarek (2021) examined ‘trustworthy information.’ Although these are certainly interesting topics to study, they do not examine trustworthiness in the sense that most philosophers of science cited above prompt us to do. That is, they do not consider trustworthiness in the normative sense, of whether an individual is deserving of trust.

Trustworthiness, in this normative sense, is about consideration of the tangible things that are necessary, in behaviour, in commitments, in material conditions or otherwise, for us to conceive of an actor as trustworthy. In the context of institutional science, this can be illustrated using the case of biobanks (O’Doherty et al. 2011). DNA biobanks are important resources for genomics research. These kinds of biobanks collect DNA samples from a very large number of people and provide these DNA samples and associated personal information about the individuals who donated those samples to researchers. Understandably, there are concerns about privacy, how best to gain appropriate consent for research, and so on. Both the biobanks and the researchers who want to use the biobank have an interest in people trusting the biobanks so that as many people as possible donate DNA and are satisfied for it to remain there. As such there is a good amount of research that asks what factors lead people to trust biobanks. The focus in much of this research is on the person doing the trusting (or not doing the trusting, as the case may be). The focus is not primarily on the practices of the biobank, but rather on the intrapsychic variables that can be manipulated to increase the trust toward biobanks inherent in any given person. The shift to trustworthiness, in this particular context, suggests that the primary focus should not be on the psychology of the people doing (or not doing) the trusting, but rather on the biobank. And the questions we should be asking are,

Given what we know about the way the biobank is set up, is it deserving of our trust? How does the biobank treat and store biological samples? How does the biobank store personal information? What kind of safeguards are in place to prevent theft of samples or data? How are researchers vetted before they are given access to biobank resources? What ethics oversight is in place to ensure that discriminatory or other objectionable research is not carried out using my sample and personal information?

In short, trustworthiness is about turning attention away from the person asked to trust and instead focusing on the person or institution asking to be trusted. It is about reflecting critically on very specific and tangible aspects of past and present conduct and institutional structures, and acting on those insights.

Trust is relational

A second problem in common conceptions underlying studies on trust is the failure to view trust as a property of relationships, not just individuals. It is worth noting here that conceiving of trust as directed toward another person or institution (as in the case of the dominant approach) does not constitute a relational ontology of trust. Such approaches still conceive of trust as an intra-psychic phenomenon since the construct is seen to be anchored within the person doing the trusting. In contrast, I argue here that to conceive of trust as relational requires consideration of the quality of the relationship between the two parties (see also Baier 1986), beyond just the psychological state of the person.

To illustrate the point, it is useful to consider the relationship between monogamous spouses. In the case of one partner’s infidelity, it is not simply the attitudes and beliefs (i.e. intrapsychic conception of trust) of the other partner that are affected. It is the relationship as a whole that has been damaged. Questions by the transgressing partner such as ‘what factors would lead to your trust in me increasing’ seem to miss the point. Any actions on the part of the transgressing partner that seek to change the attitude of the other, rather than actions based on reflection and changes in their own behaviour, seem manipulative and insincere. Past actions have diminished their trustworthiness, and this is reflected in a change in the quality of the relationship. However, it is not just the trustworthiness of the transgressing spouse that has been diminished; it is the relationship itself that would need to be rebuilt. For trust to be regained, it seems misguided to direct efforts at the intra-psychic level toward raising the ‘amount of trust’ within the cheated upon spouse; the focus on the intrapsychic levels of trust ‘inside’ the spouse who was cheated on, shifts attention not only from the trustworthiness of the cheating spouse, but also from the relationship between the two that is in need of repair for trust to be rebuilt. Not only does the behaviour of the cheating spouse need to be addressed (i.e. their trustworthiness); it is the relationship as a whole that needs to be reconfigured, especially when there is power asymmetry.

Returning to public trust in science, in cases where trust is low it is insufficient or even harmful to locate scientific attention with the entity expected to do the trusting. When the locus of measurement and intervention is primarily on the ‘public’ and supposed trust levels ‘inside’ people, our attention is diverted not only from examining the trustworthiness of particular instantiations of science, but also from what is needed to rebuild key relationships between publics and institutional science. This is well articulated by Scheman (2001) who suggests that trustworthiness of science be understood as a relationship between varied social actors (including scientists). Scheman (2001) takes this argument one step further by arguing that scientific objectivity is not possible without taking into account the social and political considerations that are necessarily a part of examining such relationships. Indeed, expecting publics to trust knowledge claims that emerge from institutions known to be inequitable (universities, corporations, government agencies) is, in Sheman’s view, irrational. Building on Sheman’s work, Grasswick (2010; 2018) further argues that publics have broader expectations of the scientific community, beyond ‘normal science,’ and that to build trustworthy relationships with publics scientists need to engage in practices that meet these expectations.

The case of COVID-19 again serves to illustrate the argument for public trust at the societal or institutional level. Warren et al. (2020) argue that for some Black communities in the US to accept vaccinations, these communities not only need to trust that the vaccines are safe and effective, they also need to accept that the organisations offering and promoting the vaccinations are trustworthy. Warren et al. (2020) and Batelaan (2021) cite well documented events that show quite clearly that mistrust of Black communities in the health care system is warranted (see also Goldenberg 2021; Whyte and Crease 2010). The point here is not to argue that Black communities should not get vaccinated; rather, the point is that conceptualising and measuring trust only as an intrapsychic variable fails to take into account that trust is also located in relationships that have a history, that exists in particular political and material conditions. In the case of US Black communities, the relationship with the healthcare system is burdened by racism and many harms to individuals and communities. It is here, in the relationship between communities and institutions, that trust is located, not just in the minds of individuals.

In our review of research on COVID-19 and trust, two studies stood out that approached trust both from a relational perspective and with attention to the trustworthiness of actors (Cheng et al. 2020; Thompkins et al. 2020). Cheng et al.’s focus was on community-based organisations in Zhejiang Province, China. They noted that these organisations’ capacity to respond to COVID-19 is largely based on previous experiences of partnering with local government to respond effectively to natural disasters. The authors thus argue that government actors should develop long-term relationships and thereby build trust with community organisations. Similarly, Thompkins et al. (2020) argue that focusing on relationships is critical in building trust between community organisations and public health institutions. Their study originally intended to convene an in-person conference with the purpose of building trustworthy relationships between public health, mental health, spiritual leaders and the African-American community. Owing to the conditions of the pandemic, the investigators instead invited pastors, public health officials, and mental health providers to give short video presentations about their experiences with COVID-19 that were then disseminated online and in church congregations. One key message of speakers was their recognition of historical harms to vulnerable communities, health disparities, stigma, and distrust of medical institutions. Across all videos, the authors noted the emphasis on relationships and trust in creating and maintaining collaborations.

In two edited volumes, Marková & Gillespie (2008; 2012) collect a series of in-depth studies of trust/distrust as embedded on social networks and historical trajectories. Collectively, these studies develop a compelling understanding of trust from sociocultural perspectives. Marková and colleagues conceive of trust as a dialogical phenomenon. Gillespie (2012) criticises experiment and survey-based studies that seek associations between variables, as being based on a mechanistic epistemology that does not take into account historical and cultural conditions and beliefs. He suggests an alternate epistemology that is dialogical, focusing not on elements but on interactions. Marková & Gillespie (2008, 2012) also show that trust is symbolic. They argue that trust cannot be studied in an isolated fashion, because as a phenomenon it is embedded in networks of other social and psychological phenomena, which lead to distinctive manifestations of trust in different contexts (2008, xviii). An illustrative example is provided by Hosking’s (2012) analysis of the ways in which trust operates in the symbolic systems of religion and nationhood. Hosking views both religions and nation states as trust-inducing symbolic systems. Examples of the latter include national education, banking, legal, and social security systems. These trust inducing systems are not only technical, but also symbolic in that they rely on key narratives for support and to foster solidarity. In this way, trust is bound up with community membership. Finally, dialogical approaches to trust allow it to be viewed as a situated phenomenon (Choi and Han 2008, 82). It is not simply that the notion of trust is perceived differently by different people; trust manifests differently across historical moments, political regimes, religious communities, cultural movements, and social institutions.

Trust should not be about manipulation

My final criticism is that when trust is approached via the common conceptualisation articulated above, when trust is conceived as a dependent variable, and research seeks to understand independent or predictor variables that explain different levels of trust in people, this is an inherently manipulative approach to engaging with persons.

Experimental and survey-based studies that view trust as a measurable intrapsychic quantity seek associations between variables. In some cases, they also vary conditions to propose additional mediating variable (Gillespie 2012). All this points to a highly mechanistic epistemology, in which ‘understanding’ of a phenomenon comes down to identifying variables that can be used to predict level of trust in an individual or group. Since many studies on public trust in science are conducted in the context of expectations that public trust in scientific institutions should be higher, the implicit goal is to identify variables that can be altered via interventions to increase public trust. The following quote illustrates this common approach:

… previous studies have highlighted numerous correlates and antecedents of trust in science and science skepticism, that could, directly or indirectly via trust in science, predict compliance with COVID-19-related prevention guidelines. Key variables among these are political conservatism (e.g. Rutjens et al., 2018; Wilgus & Travis, 2019), religiousness/religious orthodoxy (e.g. Chan, 2018; Rutjens et al., 2018), conspiracy thinking (e.g. Lewandowsky et al., 2013; Rutjens & van der Lee, 2020) and, to some extent, education level (e.g. Morgan et al., 2018). While more general and stable personality traits are less present in the previous literature on predictors of trust in science, we believe that aspects of open-mindedness, especially intellectual curiosity (i.e. the extent to which an individual has intellectual interests and enjoys thinking; Soto & John, 2017), could also play an important role in predicting both – trust in science and compliance with evidence-based prevention guidelines. (Plohl and Musil 2021, 3)

Next to the explicit goal of predicting compliance, the problem is that such variable based studies are generally also decontextualised and do not consider the richness and history of relationships in which decisions are made (Choi and Han 2008, 80). More to the point, they do not consider the political conditions under which specific communities are asked to trust authorities, whether key actor have shown themselves to be deserving of trust, and what past harms and broken promises legitimately undermine trusting relationships. Since the common conceptualisation locates trust as a quantity within individuals, such studies are oblivious to the very specific contextual factors that matter to people and that characterise their relationships.

To return to the anecdotal spousal relationship, the approach of the common conceptualisation of trust can be caricatured as follow: in the face of evident mistrust, the cheating spouse, rather than reflecting on their past actions and considering the harm they may have caused to their partner and their relationship, instead conducts systematic research as to how the other partner’s attitude to them could be changed without the cheating spouse changing their own conduct. The focus is on the intervention that will lead to a change in intrapsychic conditions. Because the focus of individualistic approaches to trust often precludes giving attention to matters of trustworthiness and the context of the relationship, there is no possibility of reflecting on what might make actors more deserving of trust. The focus is exclusively on the ‘trustor’ and what can be done to change the quantity of trust inside them.

In summary, when the focus of research is on trust as an intrapsychic variable associated with individuals, this precludes authentic dialogue between parties in a relationship. In fact, it occludes that there even is a relationship to consider. When research is built on identifying associations between variables, the self-evident application of that research is to manipulate key variables in such a way as to lead to the desired outcome – more public trust in X – in a mechanistic fashion. The problem with this approach to research on trust on any topic is that achieving aims (such as getting elected, selling more products, or increasing compliance with health guidelines) is not predicated upon engaging with people as persons, but rather by applying psychological tricks (such as the ever popular ‘nudges’ of behavioural economics) to increase the ‘trust thermometer’ presumed to reside inside people in some form or another. In contrast to questions such as ‘what can we do to be more deserving of public trust?’ questions of the form ‘what factors are effective in raising trust’ are inherently manipulative.

So what is trust, really?

In presenting criticisms of intrapsychic (or psychological) approaches to trust, I have drawn attention to two possible alternative conceptions of trust: normative approaches that focus on trustworthiness (Almassi; Irzik and Kurtulmus; John; O’Neill; Whyte & Crease), and relational approaches that conceive of trust as a property of relationships rather than individuals (Ankeny; Choi; Goldenberg; Grasswick; Hosking; Marková and Gillespie; Sheman). However, approaches that focus exclusively on trustworthiness and approaches that view trust as located exclusively in relationships also have shortcomings, and – in spite of my criticisms – intrapsychic conceptions of trust do have some merits. Jesuino (2008) points out that with everyday concepts such as trust any definitions risk reductionism. Similarly, Gillespie (2008a) observes that researchers do not have a monopoly on concept of trust, and that the ‘concept is loose in the social world’ (p.126). For this reason, I propose that research on trust should forgo monolithic definitions of trust and instead (1) recognise the diversity of meanings associated with trust in everyday interactions in a diversity of settings and (2) be pluralistic in its conception of the ontology of the phenomenon.

In this final section, I argue that pluralistic approaches to studying trust should draw on aspects of all three ontologies I have examined above. In particular, following Wittgenstein (1958) and Toulmin (1958) in their criticism of scientific approaches to psychological terms, I argue that we should not make the mistake of thinking of trust as akin to a material object of which there is some measurable quantity. Rather, a more useful approach might be to view trust as a broader phenomenon³ that is situated historically and politically, and which has at least three key manifestations:

• a normative aspect (whether someone should be trusted or not),

• and a relational aspect (the nature of the relationship between a ‘truster’ and the party they do or do not trust).

• a psychological aspect (the trust felt by a person experientially),

The focus on trustworthiness has important benefits. Increasing one’s own trustworthiness requires reflection on one’s own practices and relationship. To do this requires engagement not with cosmetic changes to myself, my brand, or my communication strategies; it requires consideration of what I do and how that affects others. Ideally, such a focus on trustworthiness would provide us with much more valuable, robust, and broadly useful institutions, goods, and public actors. The shift toward trustworthiness in this sense is a shift from the psychological to the normative. That is, we do not ask what makes a person trust or mistrust, but rather we ask what conditions need to be in place such that actors should be trusted (as discussed above, several philosophers of science have already made important contributions in this regard). However, while the normative character of the concept of trustworthiness is its strength, it is also its weakness. Because normative criteria for trustworthiness are abstract, we don’t in fact know that they are perceived as trustworthy by broader publics. For example, just because I say that I have conducted an analysis on the political platform of Barak Obama and I declare that Obama is trustworthy does not mean that people will then go on to trust Obama. To return to the biobanks example, my colleagues and I developed a theoretical argument that to be trustworthy, biobanks needed to satisfy the following 5 criteria: they needed to be representative; accountable; transparent; engage in reflective practice; and consider efforts toward sustainability (O’Doherty et al. 2011). While we maintain that our argument is sound and that we consider biobanks to be trustworthy only if they satisfy our criteria, the question of whether biobanks who follow these criteria are actually more trusted by publics is an empirical question that has to be answered for specific publics in specific contexts. Thus, while research on trust should include a focus on trustworthiness, focusing only on trustworthiness is insufficient.

The focus on trust as a quality of relationships has similarly important benefits. It allows for the consideration of political and material conditions within which particular communities are situated, and how this shapes their relationships with authorities and each other. It also allows for consideration of historical conditions and treatment of particular communities by governments. Indeed, trust is intersubjective (Gillespie 2008b). However, as individuals we also experience trust subjectively. And it is in this sense that an intrapsychic conception of trust is valuable. If one party in a relationship (say, public health agencies) is not aware that they have lost the trust of another party (say, a vulnerable minority group), asking individuals in the latter group about their feelings of trust toward the former can be a meaningful act. However, to avoid being manipulative, such an inquiry needs to be conducted with attention to the relationships between affected parties and with a genuine willingness and capacity to self-reflect and changes one’s own practices.

Conclusions

This article began by considering the role of trust in democratic governance. My purpose in doing so was to situate discussions about trust in science in political considerations. When science is understood simply as a privileged truth telling process, it is difficult to justify mistrust in science as a legitimate or even rational response. From this perspective, it is also difficult to understand why trustworthiness and relationships are relevant considerations. However, when understood as a human and social activity and, more importantly, when understood as a legitimating force in policy decisions with life and death ramifications, ‘science’ is very clearly political. Scientific institutions, scientific processes, and scientists themselves are imbued with values (Douglas 2009). These values are not homogeneous within scientific communities and certainly not across societies. And, interestingly, the observation that neither the products of science nor the expertise of scientists convey moral authority is something that is acknowledged by both scholars of science and scientists themselves (Shapin 2008, 442). Scientific knowledge and expertise have an important role in the governance of social affairs. Indeed, scientific knowledge has been highly successful as a whole, and scientific expertise is essential to good policy. However, the pronouncements of ‘science’ and scientists themselves do not stand above scrutiny. They are actors within democratic systems, and epistemic and value conflicts that have ramifications for populations as a whole cannot and should not be confined to the lab. Perhaps more importantly, even when there is strong consensus around a particular area of scientific research, this does not lead to a single inevitable policy path. The pronouncements of scientists about how science is to be interpreted and applied to address social and policy issues are situated among competing values and social priorities. As such, trust in science should not be assumed as a norm. Moreover, to be trustworthy, decisions relating to many applications of science clearly rest in the kind of contested space the requires meaningful public engagement (Burgess 2014).

In my discussion, I focused on studies that were conducted in the context of COVID-19. It is, therefore, worth emphasising that COVID scholarship is not unique in this regard. Empirical research in a range of domains (health, food, environment, marketing) and on a variety of topics (genetically modified organisms, nuclear energy, biobanks, vaccination) contains many examples of the same dominant conception of trust. My intended audience includes both theorists who work on trust and empirical researchers who conduct studies on trust. I am particularly mindful of researchers in social psychology and others who predominantly adopt social cognition approaches. By this I mean approaches to the study of human phenomena that (often implicitly) assume that they are to be found ‘inside the heads’ of people, rather than in relationships, cultural practices, or societal structures (O’Doherty and Hodgetts 2019). I also hope that my arguments will be helpful to those working in science communication and public engagement with science, whether as researchers, journalists, or practitioners. While I believe that a focus on trust is very important in these domains of research and practice, it is not the panacea it is often portrayed as, especially when conceived as simply a psychological attribute. When historical and institutional aspects of relationships are ignored, when trustworthiness is assumed rather than earned, efforts to improve trust in science may well result in further erosion of public trust in science.

My purpose in this article has been to examine the ontological foundation on which studies on trust are based. Arguably, the common ontological foundation underlying many empirical studies on trust is yet another form of the deficit model of public understandings of science (e.g. Irwin 2014; Wynne 2006). The perceived deficit is not so much about insufficient knowledge, but rather insufficient trust, which is expected to be granted to science unreflectively and with deference.

There are two important issues I have not addressed in this article. The first is that in criticising prevalent ways of measuring trust, I have not offered concrete suggestions for how trust should be assessed. Second, in emphasising the need for giving attention to relational aspects of trust, I have not outlined how more positive science/society relations might be achieved. Both of these issues have already received scholarly attention, though I hope that my analysis here provides impetus for re-examination of the former issue (how to ‘measure’ trust) and renewed energy for efforts pertaining to the latter (Burgess 2014; Irzik and Kurtulmus 2021; O’Doherty and Burgess 2013; Whyte and Crease 2010).

Acknowledgements

My gratitude to Alexandra Boeriu and Angela Delodder for assistance with literature searches of empirical articles on trust. I also want to thank members of the Discourse, Science, Publics research group at the University of Guelph, Ernst Schraube, Peter Schwartz, Alexis Fabricius, and Mylène Tantchou Dipankui for helpful comments on an earlier version of the article. My thanks go especially to two anonymous reviewers who gave exceptionally helpful and constructive feedback.

Disclosure statement

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

Additional information

Funding

The research was funded by the Social Sciences and Humanities Research Council (Canada), Grant # 1006-2019-0001 (P.I. Eric Kennedy).

Notes on contributors

Kieran C. O’Doherty

Kieran C. O’Doherty is a professor in the department of psychology at the University of Guelph, where he directs the Discourse, Science, Publics research Group. His research focuses on the social and ethical implications of science and technology. In this context, he has published on such topics as governance of health data, human tissue biobanks, the human microbiome, salmon genomics, and genetic testing. Kieran’s research also emphasises public engagement on science and technology. In this regard, he has designed and implemented public deliberations in which members of the public engage in in-depth discussion about ethical aspects of science and technology and collectively develop recommendations for policy. Kieran’s research has been funded by the Canadian Institutes of Health Research, Social Sciences and Humanities Research Council, The Office of the Privacy Commissioner of Canada, the Ontario Ministry of Research & Innovation, Genome Canada and Genome British Columbia. He is editor of the journal, Theory & Psychology.

Notes

1 Contessa (2022) distinguishes between three approaches to trust, the individual approach, the semi-social approach, and the social approach. Contessa’s argument follows a different trajectory from mine and so identifies problems with approaches to trust that are beyond the scope of my argument. In contrast to Contessa, my criticisms are focused on empirical studies of trust and follow my narrower analysis of the (implied) ontological assumptions associated with trust. The criticisms I develop below are largely of the individual and semi-social approaches, and my support is for the social approach. In this sense, my argument aligns with that of Contessa.

2 See Michell (1997) for a detailed explanation of what is entailed in positing that a given phenomenon has a quantitative structure. In particular, it is insufficient to assume quantitative structure of a phenomenon (and associated measurements and statistical treatment) based only on an observation that the phenomenon exists in degrees.

3 One anonymous reviewer made the insightful point that, following Jasanoff (2004), trust could be viewed as a co-produced object that struggles to maintain coherence.