2.1. Doxastic views

This section examines whether the doxastic views of trust have the resources to explain intuition in cases like Café. As mentioned above, doxastic views of trust posit that trusting is a species of believing. One such view argues that A trusts B to perform an action $\phi$ only if A believes that B will $\phi$ (Adler Reference Adler1994; Hieronymi Reference Hieronymi2008; McMyler Reference McMyler2011; Keren Reference Keren2014; Hardin Reference Hardin2002). On these views, trusting someone to $\phi$ is merely a special type of believing that they will $\phi$ , perhaps with some special optimism, vulnerability to betrayal, or expectations about success. On doxastic accounts of trust, the attitude that an agent adopts when they trust is the attitude of belief. Doxastic accounts support their arguments with intuitive data on the interaction between trust and belief. As Hieronymi argues:

Doxastic accounts have a special concern with testimonial trust. On these views, trusting a speaker about a proposition simply causes us to believe that proposition. Whether concerned with actions or with propositions, doxastic views of trust then are committed to the following belief norm of trust:

How then would an account committed to the BNT handle Café? The hope is that a doxastic account will have the resources to predict that Amy does not or cannot trust the barista. In Café, Amy knows p: that the barista will make a good latte. On most views of knowledge, knowing p entails believing p. This means that in Café, Amy holds the relevant belief. Because she knows p, she must therefore believe p. Thus any account of trust in which BNT was the sole sufficient condition will predict that Amy is trusting the barista. This result would run counter to our intuition of Amy failing to trust in Café.

This interpretation however takes a relatively flat-footed doxastic account for granted. After all, BNT is a mere necessary condition. Can we thus look under the hood and see what other resources a doxastic view might have for accommodating intuition? Doxastic views also have an internal tension when the belief in question has certain good-making properties, for example when a trusting belief is elevated to the level of knowledge. As Adam Carter alludes to in his work, the doxastic accounts of trust will have trouble when confronted with certain evaluative norms of beliefs. As Carter puts it:

In this passage, Carter makes reference to just the sort of incompatibility between trust and knowledge that is under consideration here. The thought is that knowledge of the outcome of the entrusted action would be a guarantee of its successful completion.Footnote ⁷ Knowledge in these cases would undermine another necessary condition of trust, namely: vulnerability to betrayal. This leads us to endorse three principles about knowledge, trust, and guarantees for a doxastic account of trust. First is a principle of knowledge and guarantees:

Finally, there is a principle which articulates the impossibility of vulnerability to betrayal when an outcome is guaranteed:

From these principles, it follows that any belief which is elevated to knowledge would render the trusting relation impossible to hold. This is so because if one knows that p, then p is guaranteed. With a guarantee that p in hand, one’s vulnerability about p is eroded and trusting someone that p is thus rendered incoherent. Satisfying GO undermines VC. This is so because of IB, if one has a guarantee that p, then one is not vulnerable to betrayal about p.

As it turns out, the combination of the principles above entails the ignorance constraint for the doxastic views. This is so because if BNT is true, then all instances of trusting are instances of a subject holding a belief that p. If VC is true, which is widely agreed in the trusting literature, then vulnerability to betrayal is a necessary condition of trust. If GO is true, then knowing that p gives one a guarantee that p is true. Finally, if IB is true then a guarantee undermines vulnerability to betrayal and by transitivity undermines trust. This means that any trusting belief on a doxastic view should remain below the threshold of knowledge, which in turn means that the doxastic accounts entail the ignorance constraint.Footnote ⁹

This section has attempted to resolve the puzzle of cases like Café for the doxastic accounts of trust. It turns out that the doxastic views have the resources to respond to intuition. A combination of principles that fall out of a commitment to trusting as believing entails the ignorance constraint. This should be taken as a mark in favor of the IC. In the subsequent sections we will see whether the non-doxastic views fare better or worse in handling intuition in the case, and have the resources to endorse the IC.

2.2. Non-doxastic views

Non-doxastic accounts of trust defend the negative thesis that trusting is not merely a special type of believing. These accounts however do not rule out that trust may sometimes amount to belief. If a non-doxastic view was combined with a doxastic element, then all of the arguments presented above would apply equally well to those views. These accounts further defend the positive theses that trusting is an affective state such as optimism (Jones Reference Jones1996), taking the participant stance (Holton Reference Holton1994), or hoping (Jones Reference Jones1996; McGeer Reference McGeer2008). In this section, we will consider an account of trusting as hoping inspired by Jones and McGeer. As Jones puts it:

If trusting can be described as a form of hoping, then we can formulate a hope norm for such accounts in the following way:

HNT: A trusts B to $\phi$ only if A hopes that B will $\phi$ .

The question is whether an account of trust that is committed to HNT has the resources to deny trust in Café. As Matthew Benton has pointed out, hope, like belief, takes propositions as its content (Benton Reference Benton, Bloser and Stahl2019). What’s more, it seems that hoping whether p is incompatible with knowing p. Benton supports this thesis by appeal to linguistic data which seem to combine hope and knowledge infelicitously:

1. (1) $\#$ I know that John is in his office, but I hope that he is not.

2. (2) $\#$ I hope that John is in his office, but I know that he is not.

3. (3) $\#$ I hope that John is in his office, and/but I know that he is

4. (4) $\#$ John is in his office, but I hope he is not.

5. (5) $\#$ I hope that John is in his office, but he is not there.

6. (6) $\#$ I hope that John is in his office, and/but he is (ibid, 5).Footnote ¹⁰

The conjunctions at (1)–(6) show us that something is very odd about a self-ascription of hope combined with a self-ascription of knowledge. While (4)–(6) do not explicitly reference John’s knowledge, simple assertions of p implicitly represent oneself as knowing p. This is so because some have argued that knowledge is the norm of assertion (Williamson Reference Williamson2000). This norm states that one must: Assert p only if one knows p (KNA) (ibid, 255). At the same time, self-ascriptions of hope seem to represent oneself as not knowing whether p.Footnote ¹¹ This leads to an apparent tension when a subject asserts both conjuncts in any sentence (1)–(6).

As Robert Gordon explains, we have this intuition about hope because of hope’s implicit reliance on an unsecured outcome. Gordon offers a more complete picture of various emotions and their relationship to knowledge. He first points out that there are some emotions which are factive. That is, there are emotions that we hold only towards what we take to be facts. This includes such emotions as pleasure, thankfulness, excitement, pride, and many more (Gordon Reference Gordon2003; Benton Reference Benton, Bloser and Stahl2019: 8). In other words, these and other emotions are knowledge-requiring (Benton Reference Benton, Bloser and Stahl2019: 8). By contrast, when one is afraid that p, worried that p, or hopes that p, then they seem to by necessity not know whether p is the case. These emotions and many others are thus knowledge-precluding (ibid). Benton provides further linguistic evidence of the knowledge-precluding and requiring emotions with constructions such as the following:

1. (7) $\#$ She got the job. I hope that she did.

Alternatively, we can see the tension when a knowledge-precluding emotion is conjoined with a knowledge-requiring emotion:

1. (8) $\#$ I hope that John is in his office, and/but I am surprised that he is.

In (7), S asserts p in the first conjunct and thus represents themselves as knowing p. In the second conjunct, S expresses the knowledge-precluding emotion of hope and thus represents themselves as not knowing whether p. In (8) S uses a knowledge-precluding emotive in the first conjunct and a knowledge-requiring emotive in the second. The contention is that one who uttered this sort of conjunction would be misusing a concept and at odds with their own assertions. This linguistic data thus establish an epistemic condition on hope: Only by being ignorant of whether p is one in the epistemic position to hope that p or to put it in the talk of norms:

The ignorance norm of hope (INH) gives us a mere necessary condition on hoping. Nonetheless, failure to abide by this norm will nullify or render our attempts at hoping moot. The interesting news is that INH and HNT together will entail the ignorance constraint.

If trust requires hope, and hope requires ignorance, then by transitivity trust requires ignorance. It turns out that the hope-derived non-doxastic accounts will join the doxastic accounts in entailing the IC. The subsequent section will address objections to the ignorance constraint and consider modifications to address these objections. We will ultimately settle on a stronger constraint involving certainty and trust-based reasons.

3.1. Trust-based reasons

Perhaps what drives the intuition in cases like Weekly Call is the sort of evidence that the subject has to support their knowledge that p. In other words, it seems to make a difference to our intuition about a failure or success of trust when the epistemic reasons for our beliefs are trust-based or non-trust-based reasons. Consider the following modified cases:

Weekly Call ${\;_1}$ and Weekly Call ${\;_2}$ take very different evidential inputs to justify Allan’s belief in p. To my mind, the intuition of a failure to trust is stronger in Weekly Call ${\;_2}$ than it is in Weekly Call ${\;_1}$ . There are thus two things to say about this intuitive result. First, these are still not cases that could be explained by the inquiry constraint. This is so because neither case says anything about Allan inquiring. We could imagine that Allan has simply left his laptop open and sees his wife’s call record by accident. Allan now knows that the call has been made, and his trust has been rendered impossible. This explanation of the Weekly Call case pair offers further support for the necessity of a constraint in addition to the QC.

Additionally, the heightened intuition in Weekly Call ${\;_2}$ may indicate that the reasons for the belief are relevant to our judgment of failed or successful trusting. When Allan’s belief is based on his wife’s stellar record of calling their daughter, he may be able to trust; but when Allan’s belief is based on his wife’s call log, he cannot trust. In response, we could supplement the IC with an account of trust-based reasons. As the comparison of the above two cases makes apparent, S having trust-based reasons for a belief in p may be compatible with knowing p and trusting that p.

A full account of trust-based reasons is beyond the scope of this argument, but we can help ourselves to an intuitive notion based upon what renders an agent worthy of trust in the first place. This should include such things as their character, general reliability, their goodwill towards the trustor, and their competence. Put simply, if the reason that A knows p is the prior trustworthiness of B, then perhaps knowing on this basis is compatible with trusting. If A knows p based on independent evidence of p regardless of B’s trustworthiness, then that knowledge is incompatible with trusting B that p.

We could thus supplement the IC to indicate that if Allan knows and trusts, then the reasons for knowledge must be of the trust-based sort:

With this modification in hand, we get the result that in Weekly Call ${\;_1}$ Allan trusts his wife; and in Weekly Call ${\;_2},$ Allan cannot trust his wife. The TBIC thus better tracks intuition about the Weekly Call case pair.

A similar distinction is drawn by Schönherr who himself leans on an intuitive reading of trust-based reasons inspired by Jones and Hardin (Schönherr Reference Schönherr2024: 1). In that work, Schönherr expresses a view of the relationship between trust-based reasons and knowledge that is importantly related to, but subtly distinct from, the thought presented here. Schönherr writes: “[…] a person who asserts that she doesn’t need to trust because she knows may express that something she takes herself to know, for reasons that have nothing to do with trust, renders reliance on her outright trusting belief unnecessary” (ibid, 16). For Schönherr then, knowing p for independent reasons makes reliance on trust unnecessary.

Schönherr makes his view clear with a case of a letter of recommendation. In the Letter case, the subject Jason asks his boss, Becca, to write him a letter of recommendation. During the interview process, Jason is shown the letter and learns that it is glowing with support. Later that night Jason’s friend asks him: “Do you trust Becca to write a good letter?” Schönherr contends that the verdict in this case is that it is possible for Jason to trust Becca to write a good letter (and perhaps even rational), it is just not required. Jason should thus say in reply to his friend that he has no need to trust Becca to write the letter (ibid, 16–17). This is so because his knowledge is based on reasons that have nothing to do with trust and are thus preventing the trust-based reasons he has about Becca from doing any work in their relationship.

What I argue for in response to the Weekly Call case pair is that trust in this situation is not unnecessary but is impossible. For Schönherr, the verdict in Letter is that Jason has no need to trust Becca, while the verdict when applying the TBIC would be that a necessary condition of trusting has been violated, thus making trusting her to write the letter impossible.

To be clear, my argument does not need to assume that trust can exist without trust-based reasons. After all, it is plausible that any situation in which a subject justifiably trusts another would involve relying on such reasons. The claim instead is that independent knowledge makes relying on such reasons impossible. Moreover, we can see from examples like Letter and Weekly Call that it is not inquiry which erodes trust in these instances. It is instead that the knowledge that p is antecedently possessed by the agent and that the agent knows p for reasons that are wholly independent of any trusting relationship.Footnote ¹⁴

3.2. Certainty and trust

A second objection comes from the thesis of fallibilism. Given that no justification for knowing that p conclusively entails that p is true (Hetherington Reference Hetherington2005), perhaps knowledge is simply too weak a state to provide the requisite guarantee which is said to undermine vulnerability. In that case, are there other constraints that might satisfy our intuitions? Consider a stronger constraint, namely a certainty constraint:

On most views, certainty is a stronger epistemic state than knowledge. In fact, if we take on board a Cartesian notion of certainty, it may be so strong as to be practically unattainable. Bob Beddor has argued that we can profitably reject this “ultra-demanding” account of certainty towards a more useful epistemic concept (Beddor Reference Beddor2020). On Beddor’s view, certainty is still more demanding than knowledge, but remains an attainable ideal. Beddor first distinguishes between psychological and epistemic certainty. The former is a matter of the strength of one’s conviction that p, and the latter indicates the strength of one’s epistemic relationship to p.

One challenge to the notion of certainty is that it could be assimilated into an account of knowledge. In other words, one might be tempted to say that the relationship expressed by certainty, either epistemic or psychological, is simply the relationship that is required for knowledge. Beddor shows that analyzing certainty in terms of knowledge in this way is too quick. This is so because of some intuitive linguistic data about knowledge and certainty (ibid, 3). Consider the following sets of sentences:

1. (9) What can we know for certain?

2. (10) What can we know?

First note that these are not identical on their most natural interpretations. Sentence (9) appears to be asking for a higher or more difficult standard than does (10). This leads to the more general conclusion that (9) is not a redundant construction. Asking for something to be known with certainty is not merely asking for it to be known. Beddor also shows that this relationship between knowledge and certainty in natural language holds up cross-linguistically (ibid).

Additionally, it is natural to ascribe knowledge in some situations while denying certainty in those very same situations. The following are some examples:

1. (11) We now know with near-certainty that Russia did this with the goal of electing Trump.

2. (12) [W]e know without certainty, but with a high degree of probability, that returns over the next 10 years or so will be very poor (ibid, 4).

If certainty was merely the epistemic relationship required for knowledge, then these constructions would sound infelicitous. These constructions do not sound infelicitous, and thus, certainty is not the status required for knowledge. We have so far seen evidence that certainty should not be understood in terms of knowledge, and that certainty is a stronger epistemic state than knowledge. But just how should we understand certainty?

Beddor suggests that we extend the framework of epistemic logic to include a certainty operator:

Certainty on this view, like knowledge, involves ruling out the possibility of error. It does so by ensuring that p holds in all the worlds that are consistent with what is epistemically certain. Moreover, certainty rules out a wider range of error possibilities than does knowledge. We thus come away from Beddor’s view with two properties of certainty. First that certainty is not to be understood in terms of knowledge, and second that certainty rules out a wider range of error possibilities than does knowledge. In other words, certainty is distinct from and stronger than knowledge. This leaves us with the question of whether a certainty constraint is the more plausible constraint on trusting. Consider linguistic evidence such as the following:

1. (13) $\#$ My wife will call our daughter, I’m certain. I trust that she will.

One might worry that it is the explicit use of the term “certain” that drives the intuition of infelicity. Without positing replacing the knowledge norm of assertion with a certainty norm,Footnote ¹⁵ we must leave ascriptions of certainty in their explicit form. Still, I argue that sentence (13) sounds at best odd and unnatural. The most likely response to this utterance is some level of confusion on the part of the hearer. We may even think that the speaker is trying to reassure themselves that the call has been made.

Additionally, there is further linguistic evidence which shows that the hope accounts of trust entail the CC. First consider ascriptions of hope paired with certainty:

1. (14) $\#$ I am certain that John is in his office, but I hope that he is/ is not.

2. (15) $\#$ I am certain that the American civil war began in 1861. I hope that the American civil war began in 1861.

Then consider ascriptions of trust in place of those very same ascriptions of hope:

1. (16) $\#$ I am certain that John is in his office, but I trust that he is/ is not.

2. (17) $\#$ I am certain that the American civil war began in 1861. I trust you that the American civil war began in 1861.

The sentences at (14–17) show that the certainty constraint is a plausible constraint. The unnatural ring of these conjunctions gives us prima facie evidence that there is something amiss when a subject claims to hope and claims to be certain. We arrive at similar results when combining certainty ascriptions with trust ascriptions. We can see that by replacing “hope” with “trust,” we get the same infelicitous results. This is an unsurprising result given that any proposition about which one is certain, one must by definition also know. The hope accounts should thus endorse the CC.

What about the doxastic accounts? Recall that we were able to pull a few intuitive commitments out of those accounts. Those commitments were guaranteed outcome (GO), vulnerability condition (VC), impossibility of betrayal (IB), and the belief norm of trust (BNT). Taken together, these commitments generated the result that the doxastic accounts entailed the IC. What will they say about the CC?

Given that we have adopted Beddor’s notion of certainty, any proposition which satisfied the epistemic certainty operator would also satisfy the knowledge operator. Recall that in that model, the $\kappa$ -alternatives are always a subset of the $\varepsilon$ -alternatives. This means that for any proposition p, if one is certain that p, then one knows p. What this means for our doxastic principles is that GO will also be satisfied for any p about which S is certain. This allows us to get the same result as with the IC. Doxastic accounts thus should endorse the CC because of the tension between trusting and securing an outcome. In case the objector thinks that knowing and trusting simply must be compatible, perhaps they could be glad that certainty can explain the intuition.

3.3. Trust-based certainty

The certainty constraint however looks similarly vulnerable to the trust-based reasons objection. Consider further modified instances of the Weekly Call case:Footnote ¹⁶

Weekly Call ${\;_3}$ and Weekly Call ${\;_4}$ mirror the original case pair. The only modification is that “knows” has been replaced with “is certain.” Once again, it seems that the intuition of a failure to trust is stronger in the case in which A has reasons which are independent of his trust in his wife, namely in Weekly Call ${\;_4}$ . To explain this phenomenon, we can again appeal to trust-based reasons for certainty. To put this in the language of necessary conditions:

Moreover, this point dovetails with prior arguments by Schönherr and Rabenberg. As Rabenberg has pointed out, inquiry is compatible with trust if that inquiry is not motivated by doubt; and that: “beliefs based on trust are also compatible with inquiry” (Rabenberg Reference Rabenberg2024: 600). Additionally, Schönherr argues trust need not erode when one inquires in order to gain certainty (Schönherr Reference Schönherr2024: 13). The arguments presented here can thus be read as complementary to both Rabenberg and Schönherr. My argument has been meant to show that we might remove the inquiry element altogether. In so doing, we see that trust is compatible with an agent being certain only if that certainty is solely held in virtue of trust-based reasons.

3.4. Appearance of knowledge or certainty

A final objection argues that the mere appearance of knowledge or of certainty might be sufficient to generate an incompatibility with trust. After all, we can imagine the same Café case from above but slightly tweaked to make it such that the subject has a Gettiered belief. Gettiered beliefs are not known and hence not known with certainty. However, it seems that the intuition of an inability of the subject to trust would persist.

If this were so, that would be a bad result for the plausibility of the TBCC. This is because it would be a case in which trust was not possible due to an epistemic feature of the agent’s life, but in which they were not certain nor were they certain for exclusively trust-based reasons.Footnote ¹⁷ The TBCC would thus fail to explain this instance of an inability to trust. Moreover, if the best candidate constraint on trust was an appearance of knowledge or a justified true belief constraint, then this would capture many instances which are captured by the TBCC, but the converse would not hold.

Despite the first impression of the case, I think we need not rush to this conclusion. The issue in this objection is that the subject has the mere appearance of knowledge or certainty. At the very most, we could say that S is psychologically certain that p in this type of case. Recall that for Beddor: “Psychological certainty is a matter of strength of conviction. A belief can be certain in this sense even if it is held for no good reason. By contrast, if a belief is epistemically certain, the believer must stand in a strong epistemic relation to its content” (Beddor Reference Beddor2020: 2).

In the Gettiered Café, then it seems that the subject is at best merely psychologically certain that p. Given that S’s belief that p has an epistemic deficiency (p is either false or has some other bad making feature), this would be a case in which a subject is able to trust, but they merely do not know this to be the case. In these situations, it would be possible for A to trust B that p, but A would not undertake such trust because of the psychological certainty they entertain towards p. There are thus two upshots from this objection. First, the TBCC should be understood to apply in terms of epistemic certainty. Mere psychological certainty (or the appearance of knowledge) only gives the false impression that the subject could not trust. Second, this result demonstrates that whether one has the ability to trust may not always be something which one is in a position to know.