Empirical and Philosophical Reflections on Trust

Abstract

A dominant claim in the philosophical literature on trust is that we should stop thinking in terms of group trustworthiness or appropriate trust in groups. In this paper, we push back against this claim by arguing that philosophical work on trust would benefit from being brought into closer contact with empirical work on the nature of trust. We consider data on reactive attitudes and moral responsibility to adjudicate on different positions in the philosophical literature on trust. An implication of our argument is that the distinction between different kinds of groups—mere groups versus institutional groups—deserves more attention than is currently recognized in the philosophical literature on trust.

In the first section of the paper, we draw some basic philosophical distinctions concerning the nature and kinds of trust. In section two, we present the positions taken by Hawley (2017), who argues against trust in groups, and Faulkner (2018), who argues in favor of trust in groups. In section three, we introduce some empirical data and suggest that, albeit tentatively, this looks to undermine Hawley's position and is compatible with Faulkner's approach. We thus suggest, on the basis of the evidence that we have available, that we have reasons to prefer the position taken by Faulkner (2018) over that taken by Hawley (2017). We end by discussing some implications for distinctions between different kinds of groups relevant for future philosophical work on trust.

1. Philosophy of Trust

Much recent philosophical work on trust flows form Baier's (1986) influential paper. In the literature that has followed, it has become reasonably commonplace to distinguish a number of related concepts. The first is that of trustworthiness. We might say that a particular individual, Mary, is trustworthy. In doing so, we are attributing a property to Mary. A second related concept is that of trusting some agent to perform a particular action: we might trust Joseph to do the shopping. This form of trust (sometimes called ‘three placed trust’) is relational in structure: it connects two individuals to an action. Semiformally, this is sometimes represented as: x trusts y to φ. See, e.g., Tallant (2019: 103). Third, we might speak very generally of simply trusting. For instance, we might say, ‘Melchior trusts Caspar’. This latter notion of trust is not the chief object of study within the philosophical literature. As Faulkner (2015: 424) has it:

Almost without exception, philosophical discussion of trust focuses on the three-place trust predicate: X trusts Y to φ (see, for instance, Baier 1986; Faulkner 2007; Hawley 2014; Hieronymi 2008; Holton 1994; Jones 1996). And then understands the two-place trust predicate—X trusts Y—derivatively.

We also want to note here that it is fairly standard within the philosophical literature to draw a distinction between trust and what is sometimes called ‘mere reliance’ (See for example, Baier 1986: 234; Hieronymi 2008: 215; Holton 1994: 2–3; Jones 1996: 14; Jones 2004: 4; McLeod 2015; Potter 2002: 3–4; Pettit 1995: 205). Cases of mere reliance fall short of involving a full, morally rich notion, and instead are just cases where we rely upon others or upon objects. For instance, though we may say that we ‘trust a shelf to hold a vase’, we do not mean to say that we would blame the shelf if it failed to hold the vase or that we would find it morally responsible for letting us down. In contrast, if we say, as we did above, that we ‘trust Joseph to do the shopping’, and he does not do the shopping, then there is plausibly a sense in which Joseph is at fault in the absence of relevant extenuating circumstances. Ordinarily, when we trust Joseph to do something and he does not, we blame him (and hold him morally responsible) for not completing the task.

Following this philosophical literature, we will endeavour to speak of ‘trust’ to signify the morally loaded notion and ‘mere reliance’ for its nonnormative counterpart, though we will have to modify this approach just a little when introducing and discussing Faulkner's work for reasons that will become clear. Throughout the remainder of the paper, we will ensure that we keep the various notions apart, referring to them explicitly as we go. With these theoretical distinctions drawn, we now want to turn our attention to trust as it applies to groups.

2. Philosophy and Trust in Groups

Our focus in this paper will be on two recent accounts and discussions of trust in groups offered by Hawley (2017) and Faulkner (2018). In short, Hawley thinks we can dispense with the notion of trust in groups. Faulkner disagrees. In this section, we'll discuss the two accounts and how they bear on the question of trust in groups. In the sections that follow this one, we will evaluate these approaches in relation to various tranches of data on trust in groups and look to the data to surface some general lessons for the philosophy of trust.

2.1 Hawley on Trust in Groups

Let us begin by getting to grips with Katherine Hawley's (2017) ‘Trustworthy Groups and Organisations’ and her account of trust in groups. Hawley asks whether the predominant interpersonal focus in the philosophical literature is appropriate, or whether we should also speak in terms of group trustworthiness and appropriate trust in groups. To be clear from the outset: Hawley maintains that we can do away with the notion of trust in groups.

Hawley begins by setting up the trust/reliance distinction. As she notes, when it comes to discussions of trust:

philosophers’ accounts are typically focused on issues of trust in individuals, aiming to capture the nature of this attitude, the circumstances under which it may be rational, and the notion of trustworthiness which corresponds to this attitude of individual trust . . . In particular, philosophers often distinguish trusting someone from merely relying upon her, then theorise this distinction. (2017: 231)

With this noted, Hawley's aim in her paper is to argue against preserving the trust/mere reliance distinction in cases involving groups. As she puts it:

I examine the costs of abandoning the trust-reliance distinction in collective contexts. Vindicating this distinction has been regarded as an essential criterion of success for accounts of trust in individuals. But, I will argue, we can explain and justify much of our practices around groups without using this distinction. (2017: 232)

Hawley then suggests that we should stop thinking in terms of either group trustworthiness or appropriate trust in groups. Instead, in her view, all of the necessary theoretical work can be done by treating trust as an interpersonal matter, and apparent trust in groups as consisting in either trusting individuals within the groups or (merely) relying on the groups themselves.

Very generally, Hawley's strategy is to identify cases where it is tempting to say that we have an instance of trust in a group, and show that, as a matter of fact, we can understand that case as one where we have an instance of trusting individuals to perform particular actions and merely relying on the group as a whole. A guiding thought in this strategy is to pursue the ‘reactive attitude’ (for instance, the attitude of blame) toward a potential breach of trust. As she notes, ‘[i]f groups are not genuinely trustworthy or untrustworthy . . . , merely reliable or unreliable, then such reactions would seem to be mistaken’ (2017: 243).

We illustrate her overarching position with just one borrowed case:

Imagine a situation in which a number of people undertake to help ensure that a group issues reliable statements. This may include members of the group, who can fill different roles in their organisation, but may also include nonmembers, for example people who founded or designed the group, or people now tasked with regulating or overseeing its activities. (2017: 244)

In such cases, and where the statements issued are not reliable, it seems tempting to simply say that if we genuinely trusted the group, then we would blame the group (as whole). Instead, however, Hawley thinks that it is appropriate to blame individuals within the groups for failing to fulfill their functions within the group. As she puts it:

Each individual takes responsibility for helping ensure that the group produces a true statement, . . . , none need take responsibility for the truth of the group's statement in the way required for assertion, and nor does the group itself take such responsibility. . . . a group statement is like a reading provided by a complex machine: we can hold the designers, operators, and maintainers responsible for ensuring that the machine provides accurate readings, without regarding any individual as asserting the content of the reading. (2017: 244)

That being so, Hawley suggests (using this case and others like it) there is no theoretical cost associated with denying the trusting/relying distinction. For Hawley, the ethical features of trust apply equally to testimony and to practical action (2017: 246). So even though the illustrated case above refers to group testimony, Hawley's point holds for debates about the moral relevance of trust to group action. We trust the individuals involved and merely rely upon the group. For her, talk of ‘trusting groups’ needs to be understood, not at face value, but instead reinterpreted as a disguised way of talking about our trust in the individuals that make up those groups. It is this position that we will seek to undermine in upcoming sections that engage with empirical studies.

2.2 Faulkner on Trust and Trust in Groups

Faulkner (2018) has trust in government as his primary focus, but since government itself is a group, and Faulkner proceeds from a general theory of the nature of trust (and trust in groups, treating government as a special case of a group), his work is of interest to us here.

Faulkner begins by setting up some distinctions. We repeat these here to facilitate the reader's engagement with what follows. The four kinds of trust that Faulkner describes are:

Predictive trust (2018: 30): this is to be understood as reliance with a belief about what the outcome will be. This is roughly synonymous with the notion of ‘mere reliance’ that we introduced above. If we describe ourselves as ‘trusting a shelf’ to hold the weight of a vase, we are relying on the shelf to take the weight of the vase and we have the belief that the shelf will hold the weight of the vase. This, for Faulkner, is predictive trust. There is no moral element to such a notion of trust.

Affective trust (2018: 30): this is to be understood as reliance with normative expectations. This is most readily understood as interpersonal trust with a moral dimension. Thus, if Alice (appropriately) trusts Verity to water her plants, and Verity does not do so, then Alice will have grounds to feel betrayed by Verity and to blame her for the subsequent death of the plants.

Both predictive and affective trust are forms of particularized trust. Both are what we introduced above as ‘three placed trust’: trust where some particular agent trusts another agent to carry out some action or actions. These can be contrasted with generalized trust of two different forms:

Generalized trust in individuals: this consists in the belief that in a range of relevant circumstances there is a (morally) right thing to do coupled to the optimistic view that others (individuals) can be relied upon to pursue that right course of action. Thus, we might say that Alice trusts Verity in this generalized sense just in case in a range of salient circumstances, Alice believes that there is a right course of action and Alice takes the optimistic view that Verity will pursue that course of action.

The group-version of generalized trust simply substitutes the optimistic view that individuals can be relied upon to pursue the right course of action for the optimistic view that groups can be relied upon to pursue the right course of action. Both of these forms of trust (generalized trust in individuals and generalized trust in government) are affective—that is, they bring with them normative expectations such that the betrayal of that trust would likely lead to the presence of reactive attitudes.

Our focus here is on this last notion of trust: group generalized trust. Faulkner maintains that we can have group generalized trust in government.

it is perfectly appropriate to talk of generalized trust of government. What is expressed here is just the idea that we can normatively expect government, and its representatives, to do the right thing—to make good decisions on our behalf, decisions that take account of our needs and dependencies—and are optimistic in holding this normative expectation. (It is unproblematic if this attributes a degree of agency to government: we happily talk of the government paying, or cutting, welfare, offering tax breaks, convicting the offender and so on . . .). (2018: 640)

Since this trust in government would be a trust in a group, this leaves Faulkner and Hawley holding different positions. And, as Faulkner (2018: 641) discusses, there seem to be a number of scenarios in which specific policies would foster this general trust in government and various downstream consequences of trust in government declining. That being so, there does seem (prima facie) that this trust in a group is doing some theoretical work.

With both of these accounts set up, we now want to explore the data. As we will see, the data do not support Hawley's position on group trustworthiness as well as they support Faulkner's position on trust in groups. Crucially, for our purpose, the empirical data on group trustworthiness suggests the need for more fine-grained theoretical distinctions between different kinds of groups

3. The Data

So far, we've been concerned with setting up the theoretical background. We now want to introduce some data. Sapienza and Zingales designed a survey that was carried out among a representative sample of 1,034 American households in 2008, with the financial crisis in full swing. Respondents were asked specifically about their trust in certain people and ‘institutions’ (2012: 124) on a scale ranging from 1 to 5, where 1 means ‘I do not trust at all’ and 5 means ‘I trust completely’. See figure 1.

Figure 1. Trust

The average response, on a scale from 1 to 5, to the question ‘How much do you trust . . . ,’ where 1 means ‘I do not trust at all’ and 5 means ‘I trust completely’ (Sapienza and Zingales 2012: 126).

Reflecting on their findings, the authors ask ‘[w]hy does trust in some types of institutions (or people) suffer compared to trust in “other people” in general?’ (2012: 124). They surmise that the average employee of those institutions ‘must be considered worse than the average person’, or, more interestingly, that ‘the institutions themselves must be thought to have negative effects on individuals,’ so much so that bankers are considered ‘less trustworthy human beings than people in general (2.60 versus 3.33)’ (2012: 125).

A striking feature of this study is its suggestion that banking institutions are more trustworthy than bankers (2.95 versus 2.60). We think that these data bring some pressure to bear on Hawley's position. Respondents trust the banks a little, but trust bankers much less. But, of course, if apparent trust in the banks is nothing more and nothing less than trust in the bankers, then these two figures should be the same—or, allowing for error, very similar. So, this is a challenge to Hawley's position. And, since Faulkner allows for trust in groups, the data seem to support his position better. To be sure, more work is needed before we can be definitive. Nonetheless, the study is noteworthy as it offers prima facie reason to prefer Faulkner's account to Hawley's.

Notice the formulation of Sapienza and Zingales's questionnaire: they tracked average responses, on a scale from 1 to 5, in response to the question ‘How much do you trust . . . ,’ where 1 means ‘I do not trust at all’ and 5 means ‘I trust completely’. Respondents were not asked whether they trust bankers, banks, the government, and large corporations to carry out a particular action or task. Rather, survey respondents were asked to choose a scale that best represents how much they trust these entities in general. Sapienza and Zingales's (2012: 124) measurement of trust is informed by Gambetta's (2000) definition of trust, where trust is ‘the expectation that another person (or institution) will perform actions that are beneficial, or at least not detrimental, to us regardless of our capacity to monitor those actions’. Notice that this definition clashes with Hawley's exclusive focus on three-placed trust and her insistence that all instances of trust in collective action and group testimony are necessarily reducible to trust placed in individual members. Gambetta's definition does not straight forwardly align with Faulkner's approach either. And although it resembles to some extent Faulkner's fourth category of trust—generalized trust in groups—it lacks the affective, normative element that is essential to Faulkner's fourth category of trust. Given that Faulkner studies generalized trust in addition to particularized trust, his approach seems better supported than Hawley's by Sapienza and Zingales's study.

3.1 The Reply

Of course, there is scope for Hawley to offer a reply. As above, we noted that there is a distinction to be drawn between trust and mere reliance. We also allowed that there are frequently occasions upon which it is perfectly reasonable for us to speak of trusting those that we are, in fact, merely relying upon. Thus, ‘I trust the shelf to take the weight of the vase’ need not be thought of as signaling that what is in play is moral trust. So, here is an imagined reply to our attempt to put pressure on Hawley. What respondents have done here is expressed their trust in bankers, but when they used the term ‘trust’ with reference to the banks themselves, the notion that they were exploring was, in fact, that of mere reliance (in Faulkner's term ‘predictive trust’). That being so, the data presented brings no pressure to bear on Hawley's position and does nothing to support Faulkner's. A variant of this reply goes like this: suppose survey respondents really did mean to say that they trusted banks as opposed to merely relied on banks. It is still possible that what respondents had in mind when they thought about the ‘bank’ was the key decision-maker in charge—the individual at the top of the hierarchy as opposed to a lower-ranked banker who works as a customer service representative.

We don't think that these lines of argument are quite right. We'll pursue two types of response. Both types of response look to undermine the thought that responses here have simply treated the banks as merely reliable. In the first case, we'll look at data that appears to show that people generally had the right kinds of reactive attitudes toward the banks to be trusting them. In the second case, we'll look at testimony that demonstrates that at least some people thought that moral responsibility, as opposed to mere reliance, lay with the banks themselves, rather than the individuals.

3.1.1 Data on Reactive Attitudes

In earlier work, Hawley is helpfully explicit on how the notions of trust and mere reliance come apart, as well as on how we might test for the presence of trust and mere reliance in particular cases. Starting with consideration of inanimate objects, Hawley locates a key difference between trust and reliance in terms of the reactive attitudes that we have in cases where it turns out that we have misplaced our trust/reliance:

Suppose I trust you to look after a precious glass vase, yet you carelessly break it. I may feel betrayed and angry; recriminations will be in order; I may demand an apology. Suppose instead that I rely on a shelf to support the vase, yet the shelf collapses, breaking the vase. I will be disappointed, perhaps upset, but it would be inappropriate to feel betrayed by the shelf, or to demand an apology from it. (2014: 2)

She goes on to illustrate how it is that these reactive attitudes can also help us understand the difference between misplaced trust and misplaced reliance.

Suppose you regularly bring too much lunch to work, and leave the excess for others to eat. Suppose you do this because you're bad at judging quantities, not because you're keen to feed your colleagues. I rely on you to provide my lunch: I anticipate that you will do so, and I don't make alternative arrangements. But this reliance should not amount to trust: you would owe me no apology if you ate all the food yourself, and I ought not to feel betrayed by this, even if I felt disappointed (and hungry). Likewise ‘Kant's neighbors who counted on his regular habits as a clock . . . might be disappointed with him if he slept in one day, but not let down by him, let alone had their trust betrayed’. (Hawley 2014: 2; citing Baier 1986, 235)

Straightforwardly, then, Hawley's contention is that when looking to distinguish cases of mere reliance from genuine trust, we ought not to look for the distinction to be consistently marked in the language that people use. Rather, we should look to the kinds of reactive attitudes that people exemplify. If we find attitudes of betrayal or anger, then we are likely dealing with a case of trust. If we do not find those attitudes, but, in fact, find disappointment (as we might in the case of a failed shelf), then we have mere reliance and not trust.

The question we should then ask is: Do we have any reason to think that there is anger at the banking industry/the banks? If such reactive attitudes are present, then that would suggest that what we have in play is not mere reliance, but trust. After all, we would have located a signifier of trust (that is, anger) rather than a signifier of mere reliance (that is, disappointment).

We believe that there is reason to think that, during and shortly after the financial crisis, there was indeed a good deal of anger at the banks and the banking sector. At a somewhat superficial level, this comes through in the kind of language used in prominent news outlets where it's not hard to find a variety of claims made about public anger at the banks, coupled to the banks being blamed for the crisis (Wilson 2011; Myners 2010; Erman 2013). But these reports, though representative of something, are not focused tightly enough for our purposes. They tend to conflate general feelings of anger with anger at banking institutions. As such, we won't make recourse to the popular press to support our contention that there was anger at the banks/banking sector. Instead, we look to the academic literature and studies that have concerned themselves with anger at the banking sector in one way, shape, or form.

Bennett and Kottasz (2012) undertook an analysis of attitudes toward UK banking. In particular, they were concerned with exploring the circumstances and characteristics that contributed to an individual's attitudes toward the banking sector in general. Two of the hypotheses that they investigated were:

H6. People who felt a great deal of anger at the banks’ role in the crisis currently tend to have less favourable evaluations of the banking industry than people who did not feel a great deal of anger at the banks’ role.

H7. People who were personally affected by the crisis tend to feel more anger at the banks’ role than people who were not personally affected. (2012: 132)

For our purposes, of course, it does not matter too much whether or not respondents ended up with negative or positive evaluations of the banks. What matters to us is whether or not people did, in fact, feel anger toward the banks. In short: they did, at least to some degree. Bennett and Kottasz found that respondents were at least somewhat angry with the banks, with anger greater in some parts of the population. Participants were asked to score the following statements on a 5-point scale (5 being strongly agree, 1 being strongly disagree).

1. (a) The behaviour of the banks during the financial crisis did not make me feel angry [reverse scored (RS)].

2. (b) I was not at all irritated or annoyed by the way the banks behaved during the crisis (RS).

3. (c) When I think about the way the banks behaved during the crisis, I feel like screaming at someone or banging on an object.

4. (d) The banks’ behaviour during the crisis makes me so mad that I feel I am about to blow up. (2012: 146)

Importantly, these questions do not ask respondents about their feeling toward individual bankers. Rather, the question asks them directly about their feelings toward ‘the banks’. Thus, at least in the first instance, answers to these questions ought to be read as delivering data about the institutions themselves, rather than the attitudes of participants toward individual bankers.

It might be objected that a study that asks individuals to choose how they feel about banks ‘primes’ them by suggesting that banks are apt targets for feelings of anger. As it happens, Bennett and Kottasz (2012) also asked a number of questions from respondents regarding trustworthiness of individuals (people in general, as well as ‘officials’ and ‘managers’). There were also additional questions regarding respondents’ attitudes toward banks, including questions about the management and leadership of the bank versus the general reliability of business and the banking industry as a whole. The broad swath of empirical data in this study included questions about individuals as well as organizations/businesses, and so we do not believe priming to be concern.

The overall finding among the 1,066 respondents was an aggregated score of 3.67 across all four questions. See Table 1:

Table 1. The Participants (Bennett and Kottasz 2012: 137)

This indicates feelings of anger toward the banks. Indeed, though somewhat tangential to our project here, Bennett and Kottasz concluded that, ‘[a]s anger represented a powerful determinant of current attitudes towards the banks it would be valuable to examine in greater depth how various dimensions of anger with the banks translate into specific attitudes and behaviours’ (141). In a related study, Kottasz and Bennett (2016: 297) found that, ‘Many of the respondents felt angry about the banks’ behaviour, and memories of the banks’ role in the crisis persisted’.

In order to theorize public attitudes toward the banking industry, Bennett and Kottasz draw on three different streams of literature: crisis management, reputation, and psychological attribution. Interestingly, these approaches all take a generalized, as opposed to-three-placed stance, on the definition of trust. The dependent variables in Bennett and Kottasz's study were ‘general’ perceptions of trust/distrust at the ‘organizational/sectoral’ level with respect to ‘competence’ (defined as ‘skills and abilities’) and ‘integrity’ (defined as ‘character, fairness, and credibility’) (134). This generalized approach is arguably appropriate given the consensus in the empirical literature regarding the collective context of the financial crisis.

The Bennett and Kottasz approach straightforwardly appears to clash with Hawley's exclusive focus on three-placed trust and her insistence that all instances of trust in collective action and group testimony are necessarily reducible to trust in individual members. And although there is a resemblance here to Faulkner's fourth category of trust—generalized trust in groups—it is not completely clear whether and how Bennett and Kottasz's affective, normative elements (‘fairness’, ‘integrity’) map onto Faulkner's notion of affective trust as a morally informed social relationship. Nonetheless since Faulkner permits both generalized and particularized instances of trust, his approach is better supported than Hawley's by the data.

Before we close this section, there are potential objections to be addressed. First, is anger really a reactive attitude? And, even if it normally is, does our understanding of what it is to be a reactive attitude pose any problems in relation to banks?

We do think that anger is a reactive attitude. We are not alone in believing so. For instance, in introducing the notion of reactive attitudes, Strawson defines them as attitudes belonging to involvement or participation with others in interpersonal human relationships, which include ‘resentment, gratitude, forgiveness, anger . . .’ (1974: 9). Strawson further characterizes reactive attitudes as ‘essentially reactions to the quality of others’ wills towards us, as manifested in their behaviour: to their good or ill will or indifference or lack of concern’ (14). As Abramson and Leite (2011) explain, reactive attitudes are thus ‘forms of interpersonally directed emotional responses’ to what Strawson calls good will and ill will. With respect to anger, Abramson and Leite (2011) elaborate:

It is commonplace that a particular reactive attitude is an appropriate response only to certain morally significant features. Anger, for instance, is typically thought to be an appropriate response to certain kinds of wrongdoing, wrongdoings which have specifically to do with what persons can reasonably demand of one another. (677)

Nonetheless, this brings us to a second concern. It is clear from Abramson and Leite's discussion of the idea of a reactive attitude that, if it is a reactive attitude, anger is only appropriate on some occasions. It is not always the case that anger is an appropriate response. For instance, if someone brings you a nice, but slightly underwhelming gift, the attitude of anger would seem to be unwarranted. Might that then raise problems? Might the idea be that only appropriate anger is an indicator of genuine trust? If the anger is misplaced, might we have a case where we do not, in fact, have trust?

We do not think so. We think that inappropriate anger might be associated with inappropriate trust. Tallant (2019: 109) gives an example of a child tricked into thinking that their blackboard will regularly and reliably communicate with them, so much so that they come to trust the blackboard. Were the trickster involved to have the blackboard mislead the child, the child will feel anger and betrayal toward the blackboard. Insofar as reactive attitudes are responses to the will of others, as Strawson would have them, the child may be perceiving the blackboard as having an ill will toward her.¹ Neither the anger nor the trust is genuinely justified, but that does not prevent them from being present.

A variation of the second objection may go like this. The Sapienza and Zingales data suggest that individuals express reactive attitudes not just in relation to individual bankers, but also in relation to banks. We have suggested that the presence of such reactive attitudes toward banks pushes back against Hawley's position that we should stop thinking in terms of group trustworthiness or appropriate trust in groups. It may be objected, however, that Hawley's theory is about whether reactive attitudes are fitting, apt, or appropriate, not about whether reactive attitudes exist. The Sapienza and Zingales data demonstrate that reactive attitudes toward banks are present. But this need not necessarily pose a problem for Hawley's position if the reactive attitudes toward banks are not in fact apt.

In response, we agree that the fittingness of reactive attitudes is important for Hawley's position as well as Faulkner's. It may well be that reactive attitudes toward banks are not appropriate, just as anger toward a shelf that fails to hold a vase is not appropriate. But we have not argued that the Sapienza and Zingales data on reactive attitudes toward banks is proof for the existence of trust. Rather, our point is that the 1,034 respondents in the authors’ survey of trust appear to believe that their reactive attitude toward banks (as opposed to individual bankers) is appropriate. We believe the expression of this belief in the data is relevant for Hawley's position, given the strong reading of her position that we should stop talking in terms of appropriate trust in groups. We also believe the respondents’ expression of this belief is noteworthy in relation to Faulkner's notion of generalized trust in groups.

In sum: Hawley's position is that apparent trust in institutions is to be interpreted as trust in the individuals within those institutions. In contrast, Faulkner's account permits us a generalized trust in groups. Since banks are groups (specifically, institutions, which are a subtype of group), the fact that levels of trust in institutions came apart from levels of trust in individuals in the case of banks and bankers during the financial crisis puts pressure on Hawley's position and seems to speak in favor of Faulkner's. One defensive move for a defender of Hawley's position would be to suggest that the initial data in fact tracked trust in individual bankers, but only reliance in banks. To put pressure on that claim, we've reported data that seem to show that a key hallmark of trust (the reactive attitude of anger) is present in people's feelings toward the banks. We suggest that this is evidence in support of Faulkner's position.

3.1.2 Testimony and the Moral Responsibility of Banks

As noted above, a key difference between trust and mere reliance (in Faulkner's terms, between affective trust and predictive trust) is that with trust, but not mere reliance, we allocate moral responsibility. Thus, if we find that respondents are happy to assign moral responsibility to the banks/banking institutions, and not to the individuals, then we have less reason to think that when responding to questions about trust in banks, people will simply be treating this as an instance of mere reliance (/predictive trust).

The data we want to consider is drawn from work by Olivia Nicol (2018). Nicol conducted interviews with 33 bankers who held various positions at a range of banks during the financial crisis. In these interviews, Nicol sought to interrogate (among other things) whoever was morally responsible for the financial crisis. It turns out that ‘banks’ is the most cited locus of moral responsibility across all interviewees (107–8). As Nicol puts it, ‘[t]hey mostly attribute responsibility to collective entities, such as “banks,” and they cannot precisely locate responsibility within banks’ (112).

In comparison to the earlier empirical studies discussed in this paper, Nicol's empirical work is noteworthy for her explicit theoretical engagement with the problem of collective responsibility. She sets out to learn the level of responsibility assigned to banks by bankers with respect to the financial crisis but makes space for examining responsibility both at the personal and the collective level. While her data suggest that the empirical subject of the study is ultimately a failure of collective responsibility, Nicol examines both individual and collective responsibility to understand the interplay between them in relation to the crisis.

Now, of course, there is a natural inclination to respond to this with the observation: ‘well they would say that, wouldn't they?’ Nonetheless, the fact remains that those outside this narrow sphere have also suggested that the banks are morally responsible or morally accountable for the crisis (Boatright 2013; Herzog 2019; Silver 2018; Sorrel 2018: esp. 22).

And, perhaps more pressingly, we must once again consider the overarching dialectic. In the early part of section 3, we presented data that appear to show that trust in banks differs from trust in bankers. If trust in banks is nothing more than trust in bankers, that result would not arise. We then considered a response: that respondents will be speaking of trust but really only meaning to talk about reliance because people don't trust institutions, they merely rely upon them. We challenged that in (3.1.1) by noting that people appear to display the right kinds of reactive attitudes toward banks to be trusting them, and what we find here (3.1.2) is that people are prepared to assign moral accountability and/or responsibility to the banks themselves. Of course, one of the key differences between trust and mere reliance is that the former, but not the latter, goes hand in hand with moral responsibility. The defender of Hawley thus really has no license simply to assert that when engaging with institutions such as banks, all we do is rely upon them and we do not really trust them. It appears to be really quite plausible that agents assign moral accountability to banks and not just bankers. This is a natural fit with Faulkner's position.

Even if we think it is unsurprising that bankers assign moral responsibility to banks, the mere fact that they do suggests that there is a group of agents that assign moral responsibility to banks. Since trust flows from moral responsibility, we have individuals who trust (rather than rely upon) groups. After all, what this case shows is that (some) people trust (some) banks—which is to say that some individuals trust some groups. And in that case, Hawley's position looks to be undermined.

An objection to our interpretation of the empirical data may go like this: perhaps it is correct that individuals assign moral responsibility to banks and are willing to trust banks. But is it possible that this is because individuals have a metaphysically unsound understanding of collective agency? In other words, is it possible that the survey respondents in the studies cited here have confused or unsound notions of trust and moral agency? In response we concede the possibility that individuals may trust banks due to an error about abstract, theoretical concepts. Nonetheless, we believe such an error is worthy of philosophical attention. Even if anger toward banks is not an apt reactive attitude in relation to banks, philosophers have reasons to take note of it when evaluating the moral significance of trust in financial interactions.

3.2 The Upshot

Now, of course, philosophical theorizing often requires that we draw fine distinctions. We are sure that there are things that a defender of Hawley's could say in response. In presenting the data that we have, we don't mean to suggest that we're intending to advance an unassailable philosophical thesis: that people clearly and definitely trust groups (as well as those individuals within them). Rather, we see our project as one of exploring the data and looking for the best fit. We see that fit as coming with Faulkner's position.

In her paper, Hawley puts matters thus:

I examine the costs of abandoning the trust-reliance distinction in collective contexts. . . . I will argue, we can explain and justify much of our practices around groups without using this distinction; the costs of abandoning it are low, as compared to the individual case. (2017: 232)

We think that there is a theoretical cost here that is potentially quite high. If, as we suggest, trust in groups (banks) can come apart from trust in individuals within those groups (bankers), then important empirical work will be blocked if we do not take seriously the idea that we can, in fact, trust groups. After all, in a very particular set of conditions (e.g., those surrounding the 2008 financial crisis) levels of trust in individuals differed from levels of trust in groups. This means that the very same background conditions can coincide with different levels of trust in groups and trust in the individuals that make up the groups. As is widely commented upon in social scientific literature (e.g., Beugelsdjik, de Grott, and van Schaik 2004; Inglehart 1990; Knack and Keefer 1997; Whiteley 2000; Zak and Knack 2001; Uslaner 2002), trust (in its moral form) plays a pivotal role in sustaining a transactional and prosperous economy. Furthermore, not only do modern market institutions rely on trust but they appear to be sustained and continue to thrive through the presence of trust. Indeed, individuals in wealthier economies are far more likely to express trust in other individuals as well as in groups, specifically in ‘economic and political institutions’ (Tonkiss 2009: 197). Since we transact both with groups and with individuals, appreciating how to foment and develop trust in both individuals and groups thus becomes of significant import. And so, if we are correct that we may, in fact, be able to trust both groups and individuals, and if the very same conditions can coincide with different levels of trust in them, then empirical exploration of how to optimally balance these levels of trust will be of significant value.

3.3 A Philosophical Implication

We've argued that the data we have on group trust in the case of trust in financial institutions favors Faulkner's position over Hawley's. But what is the appropriate signifier for trust in groups or for trust in individuals within groups? Does ‘group trust and trustworthiness’ do the job? This question is relevant because group interactions can vary in form. Consider Hawley's own understanding of what a ‘group’ can involve as in one of her examples discussed earlier in this paper:

Imagine a situation in which a number of people undertake to help ensure that a group issues reliable statements. This may include members of the group, who can fill different roles in their organisation, but may also include non-members, for example people who founded or designed the group, or people now tasked with regulating or overseeing its activities. (2017: 244)

This group interaction involves a social engagement among a number of individuals. But notice that there are also organizations involved here. Specifically, there are interorganizational group interactions, whereby individuals associate with one another, and extra-organizational group interactions whereby individual(s) guide a given organization from the outside, or where individuals are recipients of testimony or actions on the part of an organization. Implicit in Hawley's list of group interactions is a set of informal and formal social interactions (e.g., a hierarchy, a division of moral labor, and a monitoring activity). Let's call this range of organizational group interactions institutional.

Hawley uses the term ‘group’, not ‘institution’. And, in contrast, Faulkner explores the specific case of trust in government—itself an institution, rather than a mere group. Hawley's use of the term ‘group’ is in line with other philosophical work on collective agency, responsibility and belief in legal courts, scientific teams, engineering divisions, news agencies, consultancies, companies, churches, and governments (e.g., Schmitt 1994; Tollefsen 2006; Mathieson 2006; Fricker 2012; List and Pettit 2011). But, of course, groups are not (necessarily) institutions. Taking Miller's (2019) lead, we see Turner's definition of institutions as representative. According to Turner, institutions are:

a complex of positions, roles, norms and values lodged in particular types of social structures and organising relatively stable patterns of human activity with respect to fundamental problems in producing life-sustaining resources, in reproducing individuals, and in sustaining viable societal structures within a given environment. (1997: 6)

In contrast, groups can be more mundane. For example, some people may come together and decide to try to push a boulder up a hill. In so doing, each may be assigned a role (one person is responsible for ensuring the boulder does not drift to the left, another is responsible for ensuring the boulder does not drift to the right, two more are responsible for providing most of the force, and another is responsible for clearing the path). Such a group is not an institution. It lacks the norms and values Turner describes; it lacks the social structures or any real sense of a stable pattern of activity (beyond the completion of the short-lived task). Following this approach, we think of institutions as being entities like governments, hospitals, corporations, and legal systems.

Arguably, when it comes to group trustworthiness, the distinction between general group interactions and institutional group interactions matter. Consider two reasons. First, a focus on institutional trust offers a degree of specificity regarding the particular institutional dynamics (e.g., in markets as opposed to courts, in churches as opposed to hospitals) that may otherwise not be relevant to general group interactions. Second, and by extension, institutional relationships in groups can be importantly different from ordinary group interactions. A handful of friends or a bunch of children make up groups, but the structure of these interactions looks very different from bureaucratic exchanges guided by hierarchical mechanisms and overarching governance structures that we often find in those cases that involve institutions.

Tentatively, then, we view trust in institutions as a subset of trust in groups more generally but with some potential for significant differences between mere groups and institutions (that are a kind of groups). If this is correct, then trust in institutions deserves attention in addition to and in relation with trust in groups more generally.

This move to distinguish between trust in mere groups versus trust in institutions complements work by Onora O'Neill, who focuses on trust in a range of complex social and political institutional settings (see, for instance, O'Neill 2002). Daniel Weinstock similarly attends to the concept of ‘vertical trust’, which he views as taking place within formal, hierarchical structures in contexts in which agents must decide whether or not to ‘trust “faceless” institutions with which they have relations that are largely unmediated by sustained human contact’ (2013: 200). He notes:

Figuring out the conditions under which we should trust such powerful agents and institutions (indeed, figuring out whether ‘trust’ is the kind of attitude we should be aiming to cultivate), and articulating the ways in which these institutions can show themselves to be worthy of the trust of citizens should loom more largely in the agenda of political philosophers than it does, given how deeply we interact with these powerful institutions on an almost daily basis (the state, the market, the law, medicine, the media, to name just a few). (200)

If we accept O'Neill's and Weinstock's claims, then there is clearly important philosophical work to be done. To wit: is there a difference between the trust that we might put in (mere) groups, and the trust that we put in institutions? Focusing on our response to Hawley, might it actually turn out that mere groups are the kinds of entity we merely rely upon, but institutions are entities that we do in fact put our trust in, or that can appropriately betray our trust? And, shifting to Faulkner's account, might it actually turn out that we don't have generalized trust in mere groups, but only in institutions? We think that this is prima facie plausible. Institutions are often described as having a particular character. Among those characteristics, two are especially noteworthy: agency and trustworthiness. As Faulkner notes in his discussion of trust in government (2018: 640–1), we frequently assign agency to the institution of government. It is, we think, an open question as to whether we do the same (to the same degree) in the case of mere groups.

And, moving to the second noteworthy feature, trustworthiness is frequently described as being one of the most important features of institutions (see Narayan et al. 2000). Consider professional institutions, which were classically conceived of as trustworthy, and retained this character especially in the early twentieth century when professionalization first began to grow in Anglo-American markets for services. Going back to our earlier example in this paper about banking, banks are institutional forms, and have been so since their earliest historical form (Graebler 2011). Perhaps, then, it is institutions (not mere groups) that we can trust and distrust. We concede that this is highly speculative. We merely raise this here as an area where more work is required.

There is also an opportunity here to think about the nature of reactive attitudes in relation to trust in groups. Strawson (1974) casts reactive attitudes as responses to the will (good or bad) of others as revealed in their actions. With respect to group trustworthiness, an important question is whether a group can have a will, literally or figuratively. In so far as groups can only have a will figuratively, then Hawley may be correct that we trust individuals in a group as opposed to the group itself. But if it turns out that it makes sense to speak from a metaphysical, psychological, or organizational perspective about a group's will, or an institution's will, then there is room for us to push back against Hawley's view about trust in groups. It is possible that the features that separate institutions from mere groups have precisely these relevant features.²

These claims about institutions appear all the more plausible when we circle back to the definition of what it is to be an institution that we borrowed from Turner (1997). That is, an institution is:

a complex of positions, roles, norms and values lodged in particular types of social structures and organising relatively stable patterns of human activity with respect to fundamental problems in producing life-sustaining resources, in reproducing individuals, and in sustaining viable societal structures within a given environment.

If this account is correct, then institutions are stable entities. They persist. They organize stable patterns of activity. They are complex and contain within them norms and values. We think it at least plausible that where we have such an entity (a persisting, stable complex, that contains norms and values within it) we have the kind of entity that people may trust. At the very least, the question is deserving of further exploration.

To be clear, Hawley acknowledges variations in group structures and identities. Specifically, she discusses individual membership in groups that are ‘more robustly unified’, viz. groups that are ‘held together not just by some common feature of the members, but by informal and formal relations and structures within the group’ (2017: 247). With respect to professional self-regulatory bodies, for example, she notes that the structure of incentives and requirements of competence and loyalty are important. The same holds for her view about what she calls (apparently interchangeably) ‘institutional’ or ‘corporate’ culture or climate (248). Different group structures and identities merit attention for Hawley in so far as they enhance or discourage individual members’ trustworthiness. We disagree and think that trust and distrust in institutional groups deserves attention in their own right, distinctly from trust in individual members. Trudy Govier appears to agree with our suggestion, namely that we should distinguish institutions from mere groups when thinking about trust. As she puts it, groups can ‘trust or distrust’ other individuals and other groups, and ‘such attitudes can either be distributed among members or they can collectively characterize the group as a whole by virtue of its organizational structure’ (1997: 208). This suggests that beyond the individual members, a group's make-up/structure/institutional system can inform trust and distrust. Townley and Garfield (2013) also appear to support a similar position through their conception of ‘public trust’. As they put it, ‘we often individually or collectively trust, distrust, rely on, or fail or refuse to rely on such institutions or institutional players qua players’ (2013: 95). Faulkner's account of trust in groups is better suited for taking account of the distinction between mere groups and institutional groups. Moreover, the data that we have is a better fit with Faulkner's account of trust as it applies to groups. As we saw from an empirical perspective, there are good reasons to measure generalized trust in the context of the financial crisis: indeed, the consensus among finance scholars is that ‘the crisis’ cannot be straight forwardly or reliably reduced to identifiable, specific actions that are carried out by particular individuals, at least not in every instance. This is reflected in the widespread measurements of generalized trust in relation to the crisis.

A final remark regarding empirical data. Settling on the appropriate philosophical distinction between trust in mere groups and trust in institutional groups—the key philosophical implication of our work here—can have fruitful implications for future empirical work on trust. At present, the majority of empirical studies of trust that capture normativity do so by focusing on examining and measuring interpersonal trust. This is in part because social scientists, as opposed to philosophers, have played a dominant role in informing existing empirical work. An important downstream consequence of a normatively informed philosophical conception of generalized trust in institutional groups is therefore to offer theoretical fodder for a range of new empirical and experimental studies that examine whether and to what extent we can properly be said to trust and distrust institutional groups’ actions and decisions in practice.