Trust, Preemption, and Knowledge
Trust, Annette Baier once noted, is like air: it is everywhere around us, necessary for our normal functioning, and yet we rarely notice it unless it becomes “scarce or polluted” (Baier 1986, 234). The comparison between trust and air seems particularly apt when thinking of the role of trust within our cognitive lives and of the epistemological treatment of trust. Hardwig’s (1991) classic paper on the important, but neglected, epistemological role of trust thus aptly opens with a very similar comparison between trust and air (Hardwig 1991, 693).1 Indeed, although we would know very little, if at all, if we had not trusted, and if people around us were not often trustworthy (Shapin 1994; Hardwig 1991), epistemologists have been much slower than their counterparts in other philosophical fields in their recognition of the importance of trust. Thus, although John Locke, already in the early 1660s, noted that “trust is the bond of society” (2002, 213), epistemologists only started acknowledging the importance of trust in the epistemic domain relatively recently. As late as 1991, Hardwig could present his claim that “a climate of trust... is required ... to support much of our knowledge” as a novel claim, incompatible with the suppositions of most epistemologists, who view trust and knowledge as “antithetical” (Hardwig 1991, 693).
If Hardwig’s claim was indeed novel and controversial in the early 1990s, it is no longer controversial today. Few contemporary epistemologists see trust and knowledge as antithetical, and there is widespread agreement among contemporary epistemologists that we would know much less than we actually do if we did not operate within a climate of trust. In recent decades, epistemologists have therefore exhibited a growing interest in the epistemological role of trust and in developing an epistemological account of trust-based knowledge and justified belief (Faulkner 2007, 2011; Hardwig 1991; Keren 2014; McMyler 2011; Moran 2005; Zagzebski 2012).
This chapter will try to do three main things: first, it will describe the central tasks of a nonrevisionist epistemological account of trust. Such
Trust, Preemption, and Knowledge 115 an account, I will suggest, must answer two kinds of questions: the first concerns the nature of trust, and in particular, of trust when it functions as a source of knowledge and belief; the second concerns the epistemic significance of trust. I will then present the preemptionist account of trust (Keren 2007, 2014)2 and explain why it is better fitted than alternatives to perform these tasks. I will end by considering a central challenge to this preemptionist account and to its claim to be able to explain the epistemological role of trust: the challenge to explain, within the preemptionist framework, not only why trusting makes us vulnerable to others in distinctive ways but also how trust could play a positive epistemic role.
If we didn’t trust, and if others hadn’t often been trustworthy, we would know much less than we actually do know. On this, there is widespread agreement among contemporary epistemologists. Much of our knowledge thus depends upon trust.
However, that our knowledge depends upon trust might not mean that there is any need for an epistemological account of trust. For our knowledge might depend on trust in two distinct ways, and only one of them calls for an epistemological account. First, we might owe much of our knowledge to coordinated social action involving a complicated division of labor, which depends upon a climate of trust. Consider, for example, scientific research: scientists perform experiments in labs using expensive research materials, which are funded by universities, pharmaceutical companies and other funding agencies, who trust scientists to take proper care of equipment bought with the help of their funds; scientists, on their side, trust funding agencies to provide funds and salaries as promised. And this is, of course, only one node in an extensive chain of cooperation and trust: scientists hire research assistants, promising them payment, future employment, recommendation letters and so on. The latter trust the former to fulfill their promises; the former trust the latter to handle sensitive data and expensive material with care, to prepare experimental materials in accordance with preplanned protocols and so on. Thus, scientific research, like other forms of organized inquiry, and like other forms of coordinated social activity and division of labor, depends on trust (Baier 1986). Because much of our knowledge is owed to scientific inquiry and such inquiry requires trust, we would know much less than we actually do if this climate of trust did not exist.
However, as described here, trust’s contribution to the acquisition of knowledge, although highly important, is, at most, indirect and does not call for an epistemological treatment: if, as a result of an experiment she has performed, scientist S comes to believe that scientific theory T is false, then the fact that trust was required to secure the funding used to purchase experimental equipment does not seem to matter for the epistemological evaluation of S’s belief. The epistemic status of her beliefs does not depend on the source of her funding, but rather on whether the results she obtained support the conclusion that T is false.
However, trust can also play a different and much more direct role in the formation of our beliefs and in the acquisition of knowledge. This is what happens when a speaker tells us that p, and we take her word for it; when we are told that p and we form the belief that p because we trust the speaker. Suppose that after reaching the conclusion that T is false, S tells her friend, L, that T is false, and suppose that L believes S. Believing S, that is, believing a person rather than merely believing a proposition, involves trusting the person for the truth of what she says (Anscombe 1979). In such a case it does seem that the fact that L arrived at his belief by trusting S may be highly relevant to the epistemological status of L’s belief. Call the kind of trust involved in a case where one person believes another, or trusts the other for the truth of what she says, “epistemic trust.” Unlike the case of nonepistemic trust, the fact that a person’s belief is formed through epistemic trust does seem, prima facie, to be relevant to its epistemological evaluation.
Although much of our knowledge depends on both epistemic and nonepistemic trust, it is epistemic trust which Hardwig seems to have in mind when he complains about the neglect of trust by epistemologists. For even if much of our knowledge depends on a climate of nonepistemic trust, there is no reason to think that this fact calls for an epistemological treatment or may require “basic changes in epistemology and the philosophy of science” (Hardwig 1991, 694). Moreover, it is far from clear why anyone might consider nonepistemic trust and knowledge “deeply antithetical” (693). We should therefore understand Hardwig as arguing for the claim that much of our knowledge is acquired through epistemic trust. It is this claim that is standardly accepted by epistemologists today and on which I shall focus here. Call this the epistemic dependence claim.
Trust, Preemption, and Knowledge 117 epistemological accounts that deny the epistemic dependence claim and maintain that knowledge and justified belief cannot be acquired through epistemic trust are highly revisionist because our common practices and common judgments often recognize beliefs formed on epistemic trust as constituting knowledge; moreover, he suggests that because they are revisionist in this way, such accounts must be rejected.
Contemporary epistemologists have largely agreed with Hardwig, both on the claim that an epistemological account that denies that knowledge is often obtained through epistemic trust would have highly revisionist skeptical implications and on the claim that any account with such revisionist implications should be rejected. Indeed, within the contemporary epistemology of testimony, it is often assumed that it is not merely a weighty consideration against an epistemological account that it entails a revisionist form of skepticism about testimonial knowledge; rather, it is a “constraint” on any satisfactory account that it not have such skeptical implications (Fricker 1994, 1995, 2016; Rowley 2012): any account that does not satisfy this constraint can be quickly dismissed (Anscombe 1979; Weiner 2003; Fricker 2002).
Although I reject the idea that revisionist and skeptical accounts of trust and testimony can be dismissed (Keren 2019b; see also Pritchard 2004), I agree that there are methodological reasons for focusing our efforts at this stage, when the epistemological study of epistemic trust is still in its infancy, on developing nonrevisionist accounts of epistemic trust. It is only on the basis of a richer understanding of our common intuitions and practices that we can properly evaluate skeptical arguments that call for their rejection. Accordingly, in what follows, I will try to describe some of the main tasks facing nonrevisionist accounts of trust and explain why one type of account, the preemptionist account, is better fitted than the alternatives to succeed in these tasks.
In general, the task facing nonrevisionist accounts of epistemic trust can be divided into a number of distinct, but related, subtasks: first, such accounts must provide answers to questions about the nature of trust and epistemic trust: What is epistemic trust? What kind of relation must hold between two persons, S and L, and a proposition, p, for it to be the case that L believes p because she trusts S? How does this relation differ from other ways in which L might rely on S when forming the belief that pl And what makes this relation a trusting relation?
Second, the account must answer questions about the epistemic significance of epistemic trust: Can it ever be epistemically rational for L to epistemically trust speaker 5 and to believe p because of this kind of trust? Are there situations in which it would be irrational not to trust? What is the relation between the supposed fact that L’s belief that p constitutesknowledge and the fact that it was based on epistemic trust? And what kind of epistemic vulnerabilities are associated with epistemic trust?5
The challenge facing the nonrevisionist account of epistemic trust emerges from the fact that such an account must respect our commonsense intuitions with respect to both kinds of questions and that it is not clear how common judgments about the nature of trust in general, and epistemic trust in particular, can be accommodated with judgments about the epistemic significance of epistemic trust. On the one hand, there are reasons to think that trust, in general, and epistemic trust, in particular, involves holding a belief about the trusted person—either that she is trustworthy or that she will do what she is trusted to do (and in the case of epistemic trust: to tell the truth or to speak from knowledge); moreover, if epistemic trust is to play the kind of epistemic role often accorded to it, then the beliefs supposedly required for trust—call these “trusting beliefs”—must often be epistemically justified and thus sensitive to evidence. On the other hand, the relation between trust and evidence has certain features that might make it difficult to understand how, if indeed trust involves such trusting beliefs, these could be epistemically justified.
Accounts of the nature of trust can be broadly divided into two main camps, doxastic and nondoxastic, and their response to this challenge is what often underlies the division between them. On the one side are doxastic accounts of trust, which maintain that trusting a person to cP involves holding a trusting belief about her (Adler 1994; Hardin 2002; Fricker 2006; Hieronymi 2008; McMyler 2011, 2018; Keren 2014, 2019c). Indeed, according to some doxastic accounts—pure doxastic accounts—to trust a person to just is to hold a trusting belief about her (Hardin 2002; McMyler 2011). Nonpure doxastic accounts maintain that having such a belief is necessary, but not sufficient, for trust; such accounts might, for instance, maintain that trusting S to cP requires, beyond having trusting beliefs about S, also relying on 5 in certain ways (Keren 2014). In contrast, nondoxastic accounts deny that trusting 5 to 0 requires having trusting beliefs about her. Although trust may often give rise to or be accompanied by trusting beliefs, trusting beliefs are not necessary for trust (Jones 1996; McLeod 2002; Holton 1994; Faulkner 2007; Kappel 2014; Simpson 2012). Such nondoxastic accounts either maintain that trust requires some mental state other than belief—for instance, an affective state or attitude (Jones 1996) or a moral attitude towards the trusted person (Holton 1994)—or they may deny that any kind of particular mental state is necessarily involved in trusting (Kappel 2014, Simpson 2012).
The claim that the relation between trust and evidence is very different from the relation between belief and evidence and that, moreover, the relation between rational trust and evidence is very different from the relation between rational belief and evidence, is one of the main reasons that has led many philosophers to endorse nondoxastic accounts of trust
Trust, Preemption, and Knowledge 119 (Holton 1994; Jones 1996; McLeod 2002; Faulkner 2007). A number of features are usually cited in this context: the first is what is known as trust’s resistance to counter-evidence (Baker 1987; Jones 1996; Faulkner 2007). We tend to disbelieve accusations against those we trust and to interpret the behavior of those we trust favorably, disregarding evidence that could be taken as indications that they will not do what they are trusted to do; moreover, we tend not to arrive at such disbelief and such interpretations by weighing evidence for and against those accusations or for and against the claim that the trusted persons will fail to do what she is trusted to do. Indeed, seeking such evidence and weighing it seems to be incompatible with trust.
A second related feature of trust is the fact that trust is undermined by reflection on its basis: “[t]rust is a fragile plant, which may not endure inspection of its roots, even when they were, before the inspection, quite healthy” (Baier 1986, 260). Even when there is strong evidence supporting the trustworthiness of the person trusted, trust seems to be inconsistent with actively gathering this evidence.
Third, unlike belief, over which we arguably do not have direct voluntary control, it has been claimed that trust is subject to our voluntary control and that we can decide to trust for nonevidential reasons. Moreover, it has sometimes been claimed, such nonevidential reasons can make our trust rational. On the basis of these claims, several philosophers have argued for nondoxastic accounts of trust (Jones 1996; Holton 1994; Faulkner 2007; Simpson 2012).
However, there are also considerations supporting the claim that trust does require holding trusting beliefs about the trusted person (Keren 2014, 2019c; McMyler 2011, 2018; Hieronymi 2008). Indeed, denying this claim is particularly difficult when it comes to epistemic trust (Keren 2014,2019c). For when we trust a speaker who tells us that p, we invariably form the belief that p. If trusting a speaker did not require believing that she will tell us the truth or that she is trustworthy, it would be unclear why this systematic relation between trusting the speaker and believing what she says holds; moreover, it would be difficult to see how beliefs obtained on the basis of such trust could be sustained while known by the thinker to be based on trust.6
Moreover, both the belief about the speaker involved in epistemic trust and the beliefs to which epistemic trust gives rise must be epistemically justified if epistemic trust is to play the positive epistemic role we often ascribe to it. Thus, it won’t do to say that trust involves belief, but that because of its relations to evidence, trust systematically involves holding epistemically unjustified beliefs. For as noted earlier, any nonrevisionist account of trust must acknowledge that much of our knowledge is based on epistemic trust. If trust involved holding unjustified beliefs, then it would be difficult to see how beliefs based on trust could often constitute knowledge. Moreover, thinkers often explain how they knowor why their beliefs are justified by noting that their belief is based on trust; again, it would be difficult to see how thinkers could present their reasons for belief in this way, unless the beliefs involved in trusting could be rational. Relatedly, our criticism of thinkers who refuse to trust certain experts, or their epistemic superiors, seems to involve the claim that such thinkers are not epistemically justified in their beliefs, or that their belief is not responsibly formed, because these thinkers do not trust their epistemic superiors. Prima facie, these thoughts seem to presuppose that if trust involves trusting beliefs, these beliefs can be justified.
This is the challenge that all nonrevisionist accounts of trust must face. In what follows, I will illustrate how one type of account—the preemp-tionist account of trust (Keren 2007, 2014)—addresses this challenge. As I will argue, the account suggests a more promising way of addressing this challenge than the alternatives. Moreover, I will argue that the account has the resources to meet the most serious challenge raised against it, namely, the claim that the account cannot explain the positive role played by epistemic trust.