According to the Assurance Theory of testimony, in telling an audience something, a speaker offers their assurance that what is told is true, which is something like their guarantee, or promise, of truth. However, speakers also tell lies and say things they do not have the authority to back up. So why does understanding tellings to be a form of assurance explain how tellings can provide a reason for belief? This paper argues that reasons come once it is recognised that tellings are trusted. And the logic by means of which trust gives reason to believe is quite general; it applies equally to belief that is based on evidence rather than assurance. Outlining this logic requires the introduction of the idea of epistemic presumptions, whose truth plays the role of ensuring a connection between believer, justification and truth.
According to the Assurance Theory of testimony, in telling an audience something, a speaker offers their assurance that what is told is true, which is something like their guarantee, or promise, of truth (Moran 2005; Hinchman 2005a; Faulkner 2011; McMyler 2011). It is then through recognizing this intention to assure that an audience gains a distinctively testimonial reason for belief. The epistemological problem here is that assurances, like promises, can be empty. Speakers tell lies and say things they do not have the authority to back up. However, when we trust a promise, we put this possibility to one side. A trusting party will simply presume a promise is genuine and can be honoured. Similarly, when we trust a speaker for the truth, we presume their telling is truthful and that they are in a position to tell what they do. The aim of this paper is then two-fold. First, it is to argue that this presumption, constitutive of trust, is essential to the role trust plays in rationalising belief and action. It follows that a recognition of the epistemic role of this presumption is needed for a proper statement of the Assurance Theory of testimony. Second, it is to argue that the kind of epistemic role presumptions play in an Assurance Theory, they play more generally. This epistemic role has been largely overlooked as epistemological theorising has tended to focus on belief, and knowledge. The second aim is to address this over-sight through arguing for the epistemic importance of presumptions (which are different from contextual presuppositions).
The paper proceeds as follows. The first section considers the place of trust in an Assurance Theory of testimony. The second, the reasons for belief that trust provides. The third section offers an account of these reason based on the idea that certain presumptions are constitutive of trust. The fourth then shows how presumptions play a similar epistemic role in the acquisition of inductive knowledge. The fifth section then returns to the Assurance Theory of testimony and concludes.
The assurance theory of testimony
An epistemology of testimony should address two questions. First, there is a question about why we should accept testimony. Call this the question of the rationality of acceptance. Second, there is a question as to what grounds the knowledge, or warranted belief, that we acquire through accepting testimony. Call this the question of testimonial warrant. This latter question can be posed for any source of knowledge. One could similarly ask what explains our knowing things by accepting perceptual or recollective deliverances. However, the former question is unique to the epistemology of testimony as a source of knowledge and arises from the fact that testimony is delivered by other intentional agents. Given that speakers give testimony intentionally—that is, for reasons—audiences need reasons for accepting testimony as it is given. The Assurance Theory of testimony (Moran 2005; Hinchman 2005b; Faulkner 2007b; McMyler 2011; Fricker 2012) focuses on this question of the rationality of acceptance and offers a unique answer to it. Once an audience takes the ‘participant stance’ (Holton 1994), and thereby thinks of testimony as its speaker thinks of it, the intentional character of the testimony will be seen as determining a certain kind of inter-personal reason for acceptance.
To clarify this, consider practical reasons. Suppose that I decide to forgo a drink tonight. The fact that I intend not to drink tonight is evidence that I won’t. The degree to which it is evidence depends on my track record with this kind of decision. However, even if this evidence is good, my belief that I won’t drink tonight will not be based on the evidence this intention provides, but will rather follow from my intention not to drink (Marušić 2015). Now suppose I am meeting you at a bar tonight and tell you I won’t be drinking. You equally have evidence that I won’t be drinking, where this evidence is as good as my track record. Maybe I always say this, then succumb to the pleasures and pressures of the occasion. However, irrespective of whether your evidence is good or bad, if you take the participant stance you will accept what I say as it is an expression of my intention. The Assurance Theory then claims that something similar holds for testimony, or fact stating utterances. That is, when a speaker S tells an audience A that p, A can respond to S’s telling by treating it as a piece of evidence for p, where this is to view it in the same way that one views the readings of a thermometer; it is to take an ‘objectifying’ attitude (Strawson 1974). Or A can take the ‘participant stance’ and accept what S says on the basis of recognizing the intentions constitutive of telling.
The intentions constitutive of telling were first identified by Grice (1957), and can be given in the following definition.Footnote 1 S tells A that p if and only if (i) S intends that A believe that p, and (ii) S intends that A’s reason for believing that p is A’s recognition that (i). Ordinarily, tellings are assertions, though they are not necessarily so, since S might tell A something by gesture. But tellings are narrower than assertions: S might assert something to be funny or hurtful with no intention that A believe what is asserted. But in telling A that p, S does intend that A believes that p [condition (i)], and intends that A believe that p because of his, S’s, telling [condition (ii)]. The Assurance Theory then adds a further layer to this Gricean definition. In general, Moran (2005, p. 14) observed, we do not regard another’s desire that we do something as “any reason at all for complying”. So why should A regard S’s intention that she believe that p as a reason to accept what S tells her? To answer this question, Moran adds the further claim that in telling it is also true that: (iii) S “presents himself as accountable for the truth of what he says” (Moran 2005, p. 11). In this way, in telling A that p, S offers A his assurance that p. It is then A’s believing that S’s telling offers assurance which explains why A sees S’s telling, and the intentions that constitute it, as a reason for accepting what S says.
This explanation of why it is that A sees S’s telling as reason giving is not quite right. As Moran characterizes it, A’s reason for believing that p is (a) A recognizes (i), (ii) and (iii). However, what explains why A sees S’s telling as a reason for accepting what S says is not merely the fact that S presents himself as accountable for the truth of what he says, but additionally that A takes this presentation at face value. What needs to be added to (a) is: (b) A takes S to be accountable for the truth of what he says. This needs to be added because assurance need not be genuine. While we tell one another what we know, we also tell lies and we tell things where we do not have the authority needed to assume responsibility for their truth. Since tellings can be lies, lies equally satisfy conditions (i)–(iii). That is, in lying to A, it is equally true that: (i) S intends that A believe that p; (ii) S intends that A’s reason for believing that p be A’s recognition that (i); and (iii) S presents himself as accountable for the truth of what he says. But if A does take S’s lie as assurance, and thereby sees it as a reason for accepting what S says this will not merely be because (a) A recognizes S’s intentions in telling her that p but it will further be because (b) A takes S’s presentation of himself as thereby accountable at face value, and so takes S to be accountable for the truth of p.
The raises the question as to why A should take S’s presentation of himself as accountable for the truth of what he says at face value given that we tell lies and tell things that we don’t have the authority to assume responsibility for. Call this the problem of lies or empty assurance. A natural answer appeals to belief and interprets (b) as (b*): A believes that S is accountable and has the necessary authority. The problem is that adding a belief condition to the Assurance account of A’s reason for accepting what S says generates a dilemma for the Assurance Theory. If the reason for acceptance the Assurance Theory sets out to characterize is possessed only insofar as (b*) holds, then the possession of this reason becomes grounded on the facts that make (b*) true. Here there seem to be only two kinds of facts that can be appealed to. Empirical facts that give A reason to believe that S’s assurance is genuine and that S has the necessary authority to offer it. Or facts that establish that, other things being equal, A does not need to worry about the possibility of empty assurance, where these facts might be apriori (Burge 1993) or empirical (Graham 2012). The dilemma is that insofar as these are the only options, there ceases to be a distinctive assurance position: the epistemology of testimony becomes either reductive or non-reductive (Faulkner 2011, p. 143; Faulkner 2020b).
This dilemma can then be seen to be a false one, once trust is fully integrated into the Assurance Theory. In the testimonial context, the act of accepting S’s testimony to p can be best described as that of trusting S for the truth (as to whether p) or as that of believing S (when he says that p). These cases involve more than relying on S for truth because if A found out that S made a wild guess but luckily got it right, A would still feel let down by S, would feel that her trust was betrayed, even if no other harm was done. Telling the truth involves a commitment on S’s behalf to getting it right, which in turn presupposes that the capacity to take on this responsibility and discharge it. This is why the dilemma is a false one: in trusting S for the truth, A will thereby not worry about the possibility of S lying or not having the needed authority. Trust is thus the second layer that needs to be added to the basic Gricean mechanism. It is implicit in talk of seeing-as or seeing the intention a telling embodies as a reason for accepting what is said. The proposal is then to interpret condition (b) in terms of trust rather than belief. What needs to be added to condition (a) is: (b**): A trusts S for the truth. This attitude of trust is thereby an essential part of the specification of A’s reason for accepting what S says.
Trust as a reason for acceptance
On this account of the Assurance Theory, when S tells A that p, A’s reason for belief is determined by (a) A’s recognition of the intentions constitutive of S’s telling and (b**) A’s trusting S for the truth. Trust is then essential part of the specification of A’s reason for belief. To elaborate this reason, consider two cases, which are the good and bad cases familiar to epistemology. In both cases, S tells A that p and A believes S or trusts S for the truth. In the good case, S knows that p and genuinely assumes responsibility for A’s believing truly. The assurance is genuine. In the bad case, either S lies, and so does not actually assume responsibility for A believing truly, or S does not have the authority to discharge the responsibility assumed. The assurance is empty. Objectively these cases differ but subjectively, which is to say, from A’s perspective, these cases can be indistinguishable. If they are, in both cases A will have that reason that comes from recognizing S’s telling and trusting S for the truth. This raises the question of the nature of this reason; that is, how is it that trust gives A reason to accept S’s telling?
The answer given to this question depends on how trust is conceived. According to doxastic views of trust, A’s trusting S to φ either involves or entails the belief that S will φ. It is this belief that explains the willingness of A to rely on S φ-ing. Thus, in the testimonial context, trusting S for the truth involves believing that S is telling the truth (Adler 1994; Hieronymi 2008; McMyler 2011; Keren 2014; Marušić 2015).Footnote 2 On this conception, the answer to the question of how trust provides reason for belief is straightforward. If S tells A that p and A trusts S for the truth, then A will believe that S is telling the truth and thereby have a reason to believe that it is true that p.
The problem here is that condition (b**) A trusts S for the truth is now being interpreted so it is equivalent to condition (b*) A believes that S is telling the truth. This interpretation then re-animates the dilemma posed by the problem of lies or empty assurance. That is, if A’s trust involves or entails the belief that S is trustworthy, the question is raised as to what grounds A has for this belief. And here, again, the options can seem to be: either this belief is warranted on the basis of particular empirical evidence (and the epistemology becomes reductive); or there are general reasons, apriori or empirical, for taking speakers to be telling the truth when they purport to be (and the epistemology becomes non-reductive). Either way, the problem re-emerges that there ceases to be a distinctive Assurance position.
A response to this epistemological challenge is given by the idea that a belief in trustworthiness is itself a trusting belief, or one that is grounded on interpersonal facts, rather than empirical or apriori facts (see Hieronymi 2008; McMyler 2011). The idea here is that when A trusts S for the truth [so that condition (b**) is satisfied], the intentional character of S’s act of telling A that p [specified by conditions (i)–(iii)] gives A reason to believe that S is telling the truth. Here is McMyler (2011: p. 137, n.15): “[t]he attitude of trusting a person to φ itself involves believing that the person will φ, where this belief is justified by an irreducibly second-personal reason for belief. This is what makes this trusting belief different from other forms of belief—this is what makes it the case that this belief doesn’t involve the truster’s coming to her own conclusion about things.” Otherwise put: the trusting belief is not held on the basis of evidence, so does not insinuate a reductive theory of testimony, but is rather held on the basis of interpersonal facts, or A’s recognition of the intentions constitutive of S’s telling.
The problem is that this proposed basis for an audience’s trusting belief leaves the Assurance Theory with no response to the problem of lies or empty assurance. The problem, recall, starts from the fact that conditions (i)–(iii) are satisfied by lies. We tell the truth, but we also tell lies. So the mere fact that (a) A recognizes the intentions constitutive S’s telling her does not suffice to give A reason to accept what S says. What is further needed is the claim that (b) A takes at face value S’s presentation of himself as accountable; and the problem is then why A should do this. Trust is meant to speak to this problem: A’s taking S’s telling at face value is a case of trusting, which is to say that A takes S’s telling at face value because of trust. Condition (b) should be interpreted as (b**) A trusts S for the truth, and A’s reason for accepting what S says is then the conjunction of (a) and (b**). However, on the account of trust as trusting belief proposed, (b**) is equivalent to (b*) grounded on (a): A’s trusting S for the truth is a case of believing that S is telling the truth with this belief grounded on recognition of the intentions constitutive of S’s telling. But then considering the question of what reason A has for accepting what S tells her, if the reason provided by (b**) is grounded on (a), A’s reason for belief is ultimately no more than (a): it is no more than the recognition of the intentions constitutive of S’s telling. As such, this discussion of trust is an epistemically pointless addition since we are back with the Assurance Theory as originally characterised, which specified A’s reason for belief simply in terms of (a). As argued, the problem of lies or empty assurance means this specification of A’s reason is insufficient. It follows that a doxastic view of trust cannot explain how trust makes it reasonable to accept what one is told. Or, more precisely, that it cannot explain this given that it rightly rejects as untrusting any attempt to evidentially ground the belief that the speaker is telling the truth.
According to non-doxastic views of trust, A’s trusting S to φ neither involves nor entails the belief that S will φ. Thus, and for instance, trusting S to φ for Hollis (1998, p. 66) is a matter of reliance from within the “participant stance”; for Hawley (2014, 10) it is a matter of relying on S φ-ing when believing that S has a commitment to φ-ing; for myself (Faulkner 2011, p. 146; Jones 1996, p. 8), it is a matter of relying on S φ-ing and expecting S to be moved by this fact to φ. None of these further conditions—operating within the participant stance, believing the trusted to have a commitment, and expecting something of the trusted—involve or entail the belief that trusted will act in the way one relies on them acting. For non-doxastic views the question under consideration—that is, explaining how A’s trusting S for the truth makes it epistemically reasonable for A to accept what S says—is then particularly challenging. The nature of this challenge might be put like this: one can have practical reasons for trusting, which would be considerations that “show trust useful, valuable, important, or required” (Hieronymi 2008, p. 213); however, these practical reasons for trusting do not support the belief that the trusted is trustworthy. So in the testimonial case, these practical reasons do not support the belief that the speaker is telling the truth. As such, if it is also the case that the attitude of trust itself neither involves nor entails this belief, it is unclear how trust could make testimonial acceptance epistemically reasonable. In the remains of this section and the next, I describe the account I have outlined elsewhere (Faulkner 2011).
The act of trusting is one of willingly relying on someone doing something, so the general form of trust is ‘A trusts S to φ’. Essential to the identity of these trusting acts is then the trusting attitude that motivates A’s relying on S φ-ing; it is this attitude that explains A’s willingness to rely and A’s susceptibility to feelings of resentment were S to prove unreliable. On the non-doxastic account I have proposed, this trusting attitude is a normative expectation: in trusting S to φ, A expects it of S that were A to rely on S φ-ing, S would φ at least in part because A relied on him doing so (Faulkner 2007b, p. 882). Applied to the testimonial context, in trusting S for the truth, A expects S to tell her the truth as to whether p, at least in part because she relies on S for the truth as to whether p. However, one can expect things of people and also believe that they will not live up to expectations, so to explain the willingness of reliance what also needs to be observed is that trust is an essentially optimistic attitude (see Jones 1996, p. 8). Trust involves a thinking well of the trusted and holding the trusted to the expectation constitutive of trust is a way of thinking well of trusted. Thus, in trusting S to φ, or in trusting S for the truth, A takes an optimistic view of S’s motivations and competencies. This optimistic aspect of trust—this thinking well of the trusted—is expressed in A’s presumption that S would be sensitive to the reason for φ-ing, or telling the truth, given by A’s reliance on S φ-ing or for the truth; and the associated presumption that S would give this reason due deliberative weight. It follows that, other things being equal, A will presume that S is moved by this reason and so φ’s or tells the truth because of it. This presumption is then constitutive of A’s attitude of trust: without this presumption, A’s expectation is not trusting.Footnote 3
The trusting attitude is a normative expectation, where the holding of this expectation expresses a presumption, or series of presumptions, about the motivations and competencies of the trusted, and the outcome of trusting. It is then these constitutive presumptions of trust that determine the reasonableness of acts of trust. In the testimonial context, these presumptions of trust determine the reasonableness of testimonial acceptance. In the next section, I will detail how this claim fits with the Assurance Theory of testimony. But four points about these presumptions of trust should be noted now.
First, a presumption is not an explicitly held propositional attitude; it is not that A makes the presumption that S will recognize and be moved by the reason their testimonial encounter contains for S to tell the truth. However, this presumption must be attributed to A if psychological sense is to be made of A’s trust and the way this trust embeds with A’s propositional attitudes and motivates A’s actions. This is similar to how Burge (2004, p. 292) uses ‘presumption’ in his discussion’ of de se memory.
Presumption is not a propositional attitude in the individual whose states carry the presumption. A presumption that p is associated with an individual’s being in a representational state if and only if veridical recognition that p would rationally derive from fully informed, conceptually mature reflection on the conditions that make that state possible from the perspective of the individual of that state.
In these Burgean terms, states carry presumptions, which, as such, are not propositional attitudes. The suggestion is that an individual X being in state y carries the presumption that p iff X’s being in state y requires that X would recognize the truth of p if fully informed and conceptually mature. However, as Burge (2013) uses ‘presumption’ in the case discussed in the next section, the veridicality requirement is dropped: what is necessary is merely that X would judge that p if fully informed etc. But the idea remains that this requirement is needed to make psychological sense (from X’s perspective) of X being in state y. Applied to the testimonial context of S telling A that p, the state at issue is A’s expectation of S that S tells the truth because he thinks A needs the truth. The condition that makes this expectation possible, and so a presumption of A’s trust, is that S is sensitive to this reason for telling the truth and is moved by it. And if A’s trust were in place in a testimonial context where A were fully informed and conceptually mature, A would judge that this was so. This presumption is then necessary for A’s trust being the psychological state it is (from A’s perspective) since without it, A’s expectation of S would not be trusting; rather, it would be a case of expecting something of S yet anticipating non-fulfilment and resentment.
Second, insofar as presumptions are not made, they can be compared to contextual presuppositions, which, following Lewis (1996, p. 554) might be defined in terms of the ignoring of certain possibilities of truth (see also Blome-Tillmann 2014). However, given that a presupposition is a mode of ignoring, presuppositions, in contrast to presumptions, are not psychological items.
Third, given that presumptions are carried but not made, presumptions are not beliefs. And given that they are not beliefs, presumptions are not evidentially constrained in the way beliefs are evidentially constrained. Thus, A can continue, up to a point, to think well of S even in the absence of, or in the face of, the evidence (see Faulkner 2018).
Fourth, what grounds these presumptions are whatever grounds our capacity to think well of others. This will not be specific matters of evidence—since there is a tension between appeal to evidence and appeal to trust—but evidence can matter. Trust flourishes when there are social norms of truth telling and people tend to tell the truth (see Williams 2002; Faulkner 2007a); and these social facts are evidence that thinking well of people carries a tolerable risk. Arguably, we are sensitive to this evidence. But ultimately the grounds of our capacity for thinking well is not such evidence but something like an “optimistic world-view” (Uslaner 2002, p. 25) or “zest for life” (Løgstrup 1997, p. 13 and 36). And insofar as this is the ultimate grounds of these presumptions, these are not inter-personally grounded.
The presumptions of trust
Return to a case of testimony where a speaker S tells an audience A that p and A believes S or trusts S for the truth as whether p. Audience A’s recognition of S’s intentions then combines with A’s trust to give A reason to accept what S says. This reason is second-personal, or distinctively testimonial, insofar as it is crucially grounded on S’s intentions and not on the evidence S’s telling independently provides. The nature of this reason then needs to be described by reference to the good and bad cases (referred to in the last section).
In the good case, S tells A that p knowing that p and recognizing A’s need to know whether p. Moreover, this knowledge and recognition is at least part of S’s reason for telling A that p. In telling A that p, S then intends that A believe that p on his authority. In trusting S for the truth as to whether p, A expects S to tell the truth as to whether p, and do so at least in part because S recognizes her, A’s, need to know whether p. Moreover, in trusting S for the truth, A presumes that S is telling the truth, and at least in part telling the truth for this reason. Insofar as she presumes this, A will then take S’s intention that she, A, believe that p as a reason to believe that p. Audience A will thereby regard S’s telling as the assurance it purports to be and accept S’s testimony to p on S’s authority. Since this is the good case, appearances are not misleading. Speaker S does know that p and thereby has the authority to assure A of the truth of p; S is in a position to take responsibility for A believing truly. In accepting S’s testimony, and so believing that p, on the basis that S told her that p, A thereby forms the belief that p on the basis of a justifying reason for belief—a reason based, by way of the presumptions constitutive of A’s trust, on S’s decision to tell A what he knows to be the case.
In the bad case, in trusting S for the truth, A expects and presumes various things about S’s position and motivation which turn out to be false. There is no actual desire to inform or credible assumption of responsibility for A’s the presumptions of trust to latch onto. It follows that the presumptions of A’s trust can do no more than make A’s acceptance of S’s testimony to p epistemically reasonable. It is epistemically reasonable insofar as the presumptions of trust provide a reason for A to take at face value S’s presentation of himself as accountable for the truth of p. However, given that the presumptions of trust that motivate A to accept what S says are all false, and there is no actual competence or willingness on S’s behalf for the presumptions to latch on to, the reason provided for accepting S’s testimony to p is not justifying. Audience A will thereby end up with an epistemically reasonable but unjustified testimonial belief, which is probably also false.
The conjunction of A’s recognition of S’s telling and A’s trusting S for the truth—the satisfaction of conditions (a) to (b**) outlined in §1—then determines that A can have two different kinds of reason for accepting S’s testimony to p and so believing that p. In the good case, A has a justifying reason, while in the bad case A merely has a rationalizing reason. Whether A has a justifying or rationalizing reason hinges on the truth of the presumptions of A’s trust, and so on whether these presumptions get to play an anchoring epistemic role or a rationalizing epistemic role. In the good case, the epistemic role of the presumptions of A’s trust is to provide an anchor within A’s psychology of the objectively good reason for thinking that p is true that S possesses in knowing that p. That is, insofar as the presumptions of A’s trust are true and S’s possesses the epistemic authority he purports to have in telling A that p, S possesses an objectively good epistemic reason for believing p and the anchoring role of A’s presumptions of trust is that of making this reason psychologically available to A. In the bad case where A’s presumptions are false and there is no such objectively good reason to latch onto, the presumptions of A’s trust play merely an epistemically rationalizing role: they explain why A is reasonable in accepting S’s testimony to p.Footnote 4 Presumptions can play this rationalizing role (in the bad case) only because and insofar as they can play an anchoring role (in the good case). So with respect to these two epistemic roles, the anchoring role is fundamental.
The idea that reasons can be merely rationalizing and play no more role than explaining the reasonableness of action or belief is familiar (see Pettit and Smith 1990, p. 566; Moran 2001, p. 128). The idea that reasons can play an anchoring role is less familiar, and it is what I want to discuss in the remains of this paper. The hope, in doing so, is to lend further plausibility to the Assurance Theory. I start with Burge's (2013) “Postscript” discussion of his Acceptance Principle.
The Acceptance Principle states our entitlement to accept testimony (understood as utterances that intelligibly present something as true). Burge (1993, p. 467) states it thus: “A person is entitled to accept as true something that is presented as true and that is intelligible to him, unless there are stronger reasons not to do so.” Entitlement is not warrant by accessible reasons (Burge 1993, p. 458). Rather it is warrant that derives from forming belief in a way that would lead to the truth in normal conditions (see Graham 2018). Thus, the Acceptance Principle holds because, in normal conditions, accepting testimony is a good route to the truth. This claim, and the justification of the Acceptance Principle then rest on a series of apriori connections: between seeming intelligibility and rationality; between rationality and rational unity or sincerity; between being sincerely presented as true and being rationally supported or warranted; and between being warranted and being-true. Given these apriori connections—which we can call the presuppositions of the Acceptance Principle—entitlement follows because these apriori connections establish that, other things being equal, the acceptance of testimony is a good route to true belief.
The Acceptance Principle then answers what I have called the question of the rationality of acceptance; there remains the question of testimonial warrant or the question of what grounds the knowledge or warranted belief acquired through accepting testimony. Here Burge proposes that just as memory preserves knowledge and warrant across time, testimony preserves knowledge and warrant across persons. To take our simple example of S knowing that p and telling A that p, the idea is that provided A follows the minimal prescriptions of the Acceptance Principle A gets to acquire S’s knowledge. However, in discussing a slightly more complex case Burge (2013, p. 259) raises the following question:
There remains a question about the nature of the facts in [audience A’s] psycholog[y] that ground [A’s] entitlement to rely on [S] in such a way as to yield knowledge for [A]. How do[oes A] hook into … the knowledge in the antecedent chain?
Burge’s answer to this question is that two conditions must be satisfied (see Faulkner 2020a). First, the facts determining A’s entitlement to acceptance must be psychologically presumed to hold by A. That is to say, the presuppositions of the Acceptance Principle must be psychologically presumed. Second, these facts must actually hold in the particular case, so that the A’s presumptions must in fact be true.
The first requirement that the presuppositions of the Acceptance Principle be psychologically presumed is not the requirement that A reason to the truth of the Acceptance Principle or make various presumptions. Rather, it is the requirement that when ‘maturely reflecting’ on something known through testimony, A is in a position to recognize that the presuppositions of the Acceptance Principle need to hold true, where in prosaic language, these presuppositions are simply that the testimony is comprehensible, and its speaker is rational, sincere and warranted. Again, these presumptions are not made by A but carried by A’s testimonial knowledge, where this carries the presumption that p, again, “if and only if veridical recognition that p would rationally derive from fully informed, conceptually mature reflection on the conditions that make that state possible from the perspective of the individual of that state” (Burge 2004, p. 292).
The second requirement is then simply that A’s being in this knowledge state further requires these presumptions be true—that is, it must be true that A understands S’s testimony etc.. Were any of these presumptions false, A would then have an entitled belief that fails to be knowledge, or fails to be supported by preserved warranted (Burge 2013, p. 258). And if any of these presumptions were believed to be false, then other things would not be equal, and the entitlement stated by the Acceptance Principle would be defeated.
This has been a slightly long digression into Burge’s epistemology of testimony. Its point is to elaborate the anchoring role that Burge gives to presumptions. This anchoring role is to make an objectively existing warrant—the warrant that supports a speaker S knowing that p or being warranted in believing that p—psychologically available to an audience. ‘Available’ not in the sense of ‘accessible’, but available in that an audience A's presumptions secure a psychological connection to this warrant; ‘hooking in’ to this warrant and so ‘anchoring it’ in A's psychology.
In the next section I consider the case of inductive knowledge in order to argue that the anchoring role of epistemic presumptions has a wide, and widely missed, epistemic significance.
The presumptions of inductive knowledge
Suppose that one series events has been observed to correlate with another; smoke might have been observed to correlate with fire, the utterances of one speaker or kind of speaker with the truth, the readings of a thermometer with the temperature, or per capita US consumption of cheese with deaths by bedsheet strangulation.Footnote 5 That is, suppose that F events have been observed to correlate with G events. Past observations of Fs that are G can support inductive knowledge that a present F is G. So this set of observations in conjunction with a present observation of an F event allows the inductive inference to the hypothesis that there is an associated G event. The question is: when is this inductive inference knowledge supporting?
There are at least two conditions that need to be satisfied for this inductive inference to be knowledge supporting. First, there needs to be an associated G event. This follows trivially from the fact that knowledge is factive. Second, there needs to be causal connection between the F and G events, which explains the constant conjunction of F events and G events and makes it possible to know of the G event on the basis of observing the F event. Thus, and for instance, it is the causal connection between smoke and fire, thermometer readings and the temperature, and utterances and their truth, which explains the conjunction of these items and so explains why it is possible for observation of the former to ground knowledge of the latter. And it is the lack of any casual connection between cheese consumption and death by bedsheet entanglement that means that claims about the latter cannot be known on the basis of observations for the former. This requirement of a causal connection could be argued in multiple ways. It could be supported by appeal to a non-accidentality condition on knowledge (since if there were no causal connection, it would be a mere accident that there is a G event). It could be supported by appeal to a reliability condition on knowledge (since if there were no causal connection the F event would not be a reliable indicator of the G event). The argument I would like to propose appeals to a condition on something being a piece of empirical evidence.
Inductive knowledge is based on inductive evidence. So consider Achinstein (1978, p. 38) conception of potential evidence:
e is s potential evidence that h, given b, if and only if (a) e and b are true, (b) e does not entail h, (c) prob(h/e&b) > k, (d) prob(there is an explanatory connection between h and e / h&e&b) > k.
With respect to the terms of this definition, e is the presently observed F event, h is the inductively inferred G event, and b is the observed prior conjunction of F events and G events. Taking the set of observations e and b as unproblematic, condition (a) holds trivially. Condition (b) holds whenever e provides no more than an inductive basis for inferring h, which is the case by hypothesis. So the substantive conditions are (c) and (d). The probability here is objective, so the questions are respectively: does e and b render h objectively probable? That is, does the observed conjunction of F and G events and the presently observed F event render the associated G event objectively probable? And this hinges on whether or not it is objectively probable that there is an explanatory connection between the observed F event and the associated G event given the observed conjunction of F and G events? That is, it hinges on whether or not it is objectively probable that there is an explanatory connection between h and e given b? If this question can be answered positively, condition (d) is satisfied, and by implication so too is condition (c).Footnote 6 Observation e would then be potential evidence for h.
Notice, however, that condition (d) is a factual condition not a belief condition. In order for e to be evidence, certain facts must hold, it is not that these facts must be believed to hold. So it is somewhat misleading to say, as I just said, ‘if this question can be answered positively’ because there is no requirement on the subject to answer this question, condition (d) not being a belief condition. There needs, in fact, to be an objective probability of an explanatory connection between h and e given b, there is no requirement that the subject, who believes that h on the basis of e given b, also believes that there is probably such an explanatory connection.
Nevertheless, some psychological connection to the satisfaction of condition (d) is arguably necessary for inductive knowledge. Burge’s question, quoted in the last section, could be applied here. That is, how is it that the objectively good inductive warrant provided by evidence e gets to support a subject S’s belief that h? With respect to testimony, Burge argues that believing in accordance with the entitlement stated by the Acceptance Principle is not enough to inherit knowledge; there also needs to be some psychological connection to the facts that ground this entitlement. Similarly, in the inductive case, merely believing h on the basis of e given b is not enough; again, there needs to be some psychological connection to the facts that determine this basis for belief is knowledge supporting, which is to say to the facts that make e the evidence it is.
This need for a psychological connection to the facts determining that e is the evidence it is can be supported by considering a subject S* who believes that h on the basis of e given b but who follows this pattern of belief for every instantiation of F and G. That is, while S* will believe there is a fire on the basis of an observation of smoke given the past conjunction of smoke and fire, this subject will also believe there will be rise in deaths by bed sheet entanglement on the basis of a rise in cheese consumption given the past conjunction of these events. S* will project every observed conjunction. However, not all conjunctions support projection. The observation of a conjunction of cheese consumption and bed sheet entanglement does not allow the projection of bed sheet entanglement; and, famously, Jackson (1975) argues that the observation of grue emeralds does not allow the projection of grueness, even though it is coincident with observed greenness, which can be projected. But, by hypothesis, subject S* simply fails to discriminate in any way between those properties that can be projected and those that cannot. Given S*’s lack of discrimination, it seems that S* is not sensitive to when an observation e is evidence. So suppose that S* makes an inductive inference to h (a G event) from e (a F event) given b (a constant conjunction of F and G events) and suppose condition (d) is satisfied so that e is potential evidence that h. It follows that while S* believes that h on the basis of an observation e, which is in fact potential evidence, there is a sense in which it is true to say that S* does not believe that h because of the evidence e provides. But if this is true, it seems that S* is not in a position to know that h. Rather, S*’s position seems to be parallel to the subject who has an unjustified but justifiable belief (see Ginet 1975, p. 32).Footnote 7 That is, S*’s belief is clearly justifiable given the evidence S* possesses, but it is not justified, at least in the way needed for knowledge, given S*’s belief does not bear the right psychological connection to this evidence.
So what psychological connection to the evidence determining facts is needed for some subject S to know that h? The key evidence determining fact is the holding of the objective probability that satisfies condition (d); it is the fact that it is objectively probable that there is an explanatory connection between h and e given b. Call this a presupposition of S’s evidence. Arguably the belief that this presupposition is true would be a sufficient psychological connection. That is, if S believed that h on the basis of e, given b, and believed this presupposition to be true, there is no sense in which it is true to say that S does not believe that h because of the evidence. But if S believes that h because of the evidence, then S inductively knows that h when h and this presupposition are in fact true. However, while sufficient to psychologically connect S’s evidence that e to S’s belief that h in a knowledge supporting way, the belief that this presupposition is true is not necessary for this psychological connection. As Hume (1740, §1.4.1, p. 183) argued when discussing induction, “belief is more properly an act of the sensitive, than of the cognitive part of our natures”. Inductive beliefs are formed habitually often without reflection. Moreover, we allow that children and non-reflective adults can know things inductively, but it is not plausible to suppose either possess any belief about objective probability or explanatory connections.
So some psychological connection is needed to the presupposition of S’s evidence, but belief is too demanding. In this case, presumption is a natural alternative. On this proposal, in forming the inductive belief that h on the basis of e given b, S presumes that there is an objective probability of an explanatory connection between h and e given b. This presumption is not a propositional attitude, it is not made by S but carried by S’s inductively grounded belief that h. When this presumption is true, it then plays an anchoring role: it anchors within S’s psychology the objectively good epistemic reason for believing that h provided by e given b. And when this presumption is false, it plays a rationalizing epistemic role: it ensures that S is being epistemically reasonable in believing that h on the basis of e given b. This presumption is then a psychological state in that it needs to be attributed if sense is to be made of S inductively knowing that h and sense is to be made of the way that S’s inductive belief that h embeds with S’s other propositional attitudes and disposes S to think and act in various ways.
Conclusion: assurance and evidence
Return to our simple testimonial encounter, a conversation where a speaker S tells an audience A that p. In this context, audience A can think about S’s telling in two contrasting ways. It is possible for A to take the participant stance and enter into the spirit of the conversation. Since S tells that p with the intention that A believe that p on his authority in entering into the spirit on the conversation, A will take these intentions at face value and trust S for the truth. And it is possible for A to take a more objective, and objectifying, attitude, to distance himself from the conversation and consider whether S’s telling that p is evidence or not for p. In the first case, A will accept what S says on the basis of recognizing the intentions constitutive of telling; A’s reason will be S’s assurance that p is true and that he trusts S for the truth. In the second case, A will accept what S says on the basis of taking S’s telling to be evidence for p. It follows that there is a sharp contrast between treating testimony as assurance and treating testimony as evidence, and this has prompted the worry that assurance, unlike evidence, does not really provide an epistemic reason at all (Lackey 2008, ch.8; Schmitt 2010; Goldberg 2020, ch.4). This worry is one I have tried to address elsewhere (Faulkner 2011, ch.6). Here my concern has been, first, to show the mechanics of how trust and assurance combine to give audience A reason to accept what S says, and so believe that p. And my concern has been, second, to show how this reason, once laid out, is not so different to the reason provided by evidence in that both support knowledge acquisition only once certain epistemic presumptions are recognised as having an anchoring epistemic role.Footnote 8
Hinchman (2005b, p. 578) says, “[t]rust is not belief, although it may give rise to belief.”
The place of therapeutic trust, see Horsburgh (1960), is interesting here: trust can move the trusted to trustworthiness, but it remains true that the trusting party must be optimistic about this possibility for this possibility to be real. See Elster (2007, p. 350) for some good discussion of this point.
A further way in which a bad case could be bad is if trust is misplaced. Maybe the telling is so obviously incompetent it ought to be recognised as such. Thus, the notion of a rationalizing reason is thin; it is to say, that from the audience’s subjective point of view the balance of reasons suggest the testimony is true.
<https://tylervigen.com/view_correlation?id=7>. Accessed 26 June 2020.
Given the satisfaction of (c) hinges on the satisfaction of (d) it is no surprise that Achinstein (2001), changes his definition of potential evidence by dropping condition (c).
Also put by Feldman and Conee (1985) as the distinction between justification and well-foundedness.
Thanks are owed to Adam Carter and Chris Kelp.
Achinstein, P. (1978). Concepts of evidence. Mind, 87(1), 22–45.
Achinstein, P. (2001). The book of evidence. Oxford: OUP.
Adler, J. E. (1994). Testimony, trust, knowing. Journal of Philosophy, 91(5), 264–275.
Blome-Tillmann, M. (2014). Knowledge and presuppositions. Oxford: Oxford University Press.
Burge, T. (1993). Content preservation. Philosophical Review, 102(4), 457–488.
Burge, T. (2004). Memory and persons. Philosophical Review, 112, 289–337.
Burge, T. (2013). Postscript: ‘Content preservation.’ In T. Burge (Ed.), Cognition through understanding: Self-knowledge, interlocution, reasoning, reflection. Oxford: Oxford University Press.
Elster, J. (2007). Explaining social behaviour: More nuts and bolts for the social sciences. Cambridge: Cambridge University Press.
Faulkner, P. (2007a). A genealogy of trust. Episteme, 4(3), 305–321.
Faulkner, P. (2007b). On telling and trusting. Mind, 116(464), 875–902.
Faulkner, P. (2011). Knowledge on trust. Oxford: Oxford University Press.
Faulkner, P. (2018). Giving the benefit of the doubt. International Journal of Philosophical Studies, 26(2), 139–155.
Faulkner, P. (2020a). The testimonial preservation of warrant. In S. Wright & S. Goldberg (Eds.), Memory and testimony: New essays in epistemology. Oxford: Oxford University Press.
Faulkner, P. (2020b). Testimony and trust. In Simon, J. (Ed.) Routledge handbook on trust and philosophy.
Feldman, R., & Conee, E. (1985). Evidentialism. Philosophical Studies, 48(1), 15–34.
Fricker, M. (2012). Group testimony? The making of a collective good informant. Philosophy and Phenomenological Research, 84(2), 249–276.
Ginet, C. (1975). Knowledge, perception, and memory. Dordrecht: Reidel.
Goldberg, S. (2020). Conversational pressure: Normativity in speech exchanges. Oxford: Oxford University Press.
Graham, P. J. (2012). Epistemic entitlement. Noûs, 46(3), 449–482.
Graham, P. J. (2018). Sincerity and the reliability of testimony: burge on the a priori basis of testimonial entitlement. In E. Michaelson & A. Stokke (Eds.), Lying: Language, knowledge, ethics and politics. Oxford: Oxford University Press.
Grice, P. (1957). Meaning. In P. Grice (Ed.), Studies in the way of words. Cambridge: Harvard University Press.
Grice, P. (1980). Meaning revisited. In P. Grice (Ed.), Studies in the way of words. Cambridge: Harvard University Press.
Hawley, K. (2014). Trust, distrust and commitment. Nous, 48(1), 1–20.
Hieronymi, P. (2008). The reasons of trust. Australasian Journal of Philosophy, 86(2), 213–236.
Hinchman, E. (2005a). Telling as inviting to trust. Philosophy and Phenomenological Research, 70(3), 562–587.
Hinchman, E. S. (2005b). Telling as inviting to trust. Philosophy and Phenomenological Research, 70(3), 562–587.
Hollis, M. (1998). Trust within reason. Cambridge: Cambridge University Press.
Holton, R. (1994). Deciding to trust, coming to believe. Australasian Journal of Philosophy, 72(1), 63–76.
Horsburgh, H. J. N. (1960). The ethics of trust. Philosophical Quarterly, 10(41), 343–354.
Hume, D. (1740). L. A. Selby-Bigge (Ed.), A Treatise of human nature. Oxford: Clarendon Press.
Jackson, F. (1975). Grue. Journal of Philosophy, 72, 113–131.
Jones, K. (1996). Trust as an affective attitude. Ethics, 107(1), 4–25.
Keren, A. (2014). Trust and belief: A preemptive reasons account. Synthese, 191(12), 2593–2615.
Lackey, J. (2008). Learning from words—Testimony as a source of knowledge. Oxford: Oxford University Press.
Lewis, D. (1996). Elusive knowledge. Australasian Journal of Philosophy, 74, 549–567.
Løgstrup, K. E. (1997). The ethical demand. Notre Dame: University of Notre Dame Press.
Marušić, B. (2015). Evidence and agency: Norms of belief for promising and resolving. Oxford: Oxford University Press.
McMyler, B. (2011). Testimony. Trust and Authority: Oxford University Press.
Moran, R. (2005). Getting told and being believed. Philosophers’ Imprint, 5(5), 1–29.
Moran, R. (2001). Authority and estrangement: An essay on self-knowledge. Princeton: Princeton University Press.
Pettit, P., & Smith, M. (1990). Backgrounding desire. The Philosophical Review, 94(4), 565–592.
Schiffer, S. R. (1972). Meaning. Oxford: Clarendon Press.
Schmitt, F. (2010). The assurance view of testimony. In A. Haddock, A. Millar, & D. Pritchard (Eds.), Social epistemology. Oxford: Oxford University Press.
Uslaner, E. M. (2002). The moral foundations of trust. Cambridge: Cambridge University Press.
Williams, B. (2002). Truth and truthfulness. Princeton: Princeton University Press.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Faulkner, P. The presumption of assurance. Synthese (2021). https://doi.org/10.1007/s11229-021-03074-y