On behalf of a bi-level account of trust

Abstract

A bi-level account of trust is developed and defended, one with relevance in ethics as well as epistemology. The proposed account of trust—on which trusting is modelled within a virtue-theoretic framework as a performance-type with an aim—distinguishes between two distinct levels of trust, apt and convictive, that take us beyond previous assessments of its nature, value, and relationship to risk assessment. While Sosa (A virtue epistemology: apt belief and reflective knowledge, volume I, Oxford University Press, Oxford, 2009; Judgment and agency, Oxford University Press, Oxford, 2015; Epistemology, University Press, Princeton, 2017), in particular, has shown how a performance normativity model may be fruitfully applied to belief, my objective is to apply this kind of model in a novel and principled way to trust. I conclude by outlining some of the key advantages of the performance-theoretic bi-level account of trust defended over more traditional univocal proposals.

1 .

A helpful—and thus far, unexplored—way to think about trust will be to draw from the resources of virtue epistemology, and in particular, from the bi-level virtue epistemology framework developed by Ernest Sosa (2009, 2015, 2017).

Sosa’s distinction between animal knowledge and reflective knowledge (a distinction around which his epistemological project is centred) is controversial.¹ Nonetheless, his sophisticated framework for modelling the normative structure of performances with aims will be useful—regardless of what we say about knowledge—for theorising about two very distinctive levels of trust, levels that take us beyond previous assessments of its nature, value and relationship to risk assessment.

The bi-level account of trust I aim to propose is one in which the notion of competence plays a starring role. One may initially wonder why competence should play any role in an account of trust (as opposed to trustworthiness²)? Competences, after all, are dispositions, but trust is paradigmatically an attitude or a hybrid of attitudes (e.g., optimism, hope, belief, etc.), one that involves actual, and not merely dispositional, vulnerability to betrayal.

In what follows, I’ll focus (i) first on a simple auxiliary role for competence to play in an account of trust, and then (ii) on some more sophisticated constitutive roles. What results will be a novel performance-theoretic account of the structure of trust, one that will have applications in both ethics and epistemology as well as advantages over more traditional accounts in both areas.

2 .

Beliefs aim at truth.³ Sometimes, beliefs hit that aim through dumb luck. Such beliefs, while valuable to the extent that truth is valuable, are regarded as less valuable, less epistemologically praiseworthy, as knowledge—viz., what you get when belief hits its aim (truth) through cognitive skill. This is an idea that lies at the heart of contemporary virtue epistemology.

Does trust aim at anything? If it does, then a tempting line of thought goes as follows: when trust hits its aim—whatever that is—that is good to the extent that trust’s aim is a worthy aim.⁴ But, when trust hits its aim skilfully, that is something even better.

Plausibly, trust does have an aim, much like belief does. Even among those who disagree about the nature of trust, it is generally agreed that trust is an attitude we have towards people whom we hope will (in some suitably specified way) take care of things as we have entrusted them,⁵ where the entrusting itself involves incurring some non-negligible level of risk.⁶

There is much discussion about what it would be for the trustee to take care of things as entrusted—viz., what additional attitudes and/or beliefs⁷ this may involve, including attitudes and beliefs about the attitudes (e.g., goodwill) and beliefs of the trustee.⁸ However these details are filled out, it remains that when one trusts another to take care of things as entrusted, there is a clear sense in which the trust succeeds only when the trustee actually does take care of things as they are entrusted (perhaps, compresently, with certain attitudes or beliefs).

I want to accordingly take as a starting point a distinction between (i) trust as a success-apt attitude and (ii) successful trust, as follows:

Trust attitude (Trust-A): an attitude we have towards people whom we hope will take care of things as we have entrusted them,⁹ and which involves incurring some non-negligible level of risk.

Successful trust: when one’s Trust-A in another is fulfilled—viz., when another has taken care of things as we have entrusted them.

3 .

Just as one can fail to be trustworthy, one can fail to trust well, by trusting (and incurring some risk of betrayal) in ways that don’t ordinary lead to successful trust.

For example, one might be overly naive or overly cynical with respect to others’ intentions or abilities.¹⁰ In the former case, one fails to trust well if one (for instance), overestimates others’ goodwill, or if one too easily perceives bad will when it is not present. In the latter case, one fails to trust well if (for instance) one, overly jaded, too easily perceives others as incapable of taking care of things that they would have easily done.

Likewise, one might have (i) a distorted introspection of personal risk; or (ii) a distorted perception of gains of betrayal to trustee. In the former case, one fails to trust well if one (for instance) fails to appreciate what is at stake for oneself—viz., the costs to oneself of betrayal by the trustee, regardless of what the trustee stands to gain through betrayal. In the latter case, one fails to trust well if one (for instance) fails to appreciate the extent to which the trustee would benefit by betraying, regardless of the costs to the truster of betrayal.¹¹

4 .

Successful trust is compatible with failing to trust well. One might, by exhibiting any of the above (or other) defects in trusting—trust in a way that easily would not lead to successful trust, but which does anyhow on a particular occasion.

That you can trust badly and still trust successfully is analogous to what happens in other domains of performance.¹² Even the poorest chess player’s strategy might on occasion win a game.

If trusting well were entailed by trusting successfully, there would be no obvious need for an account of what it is to trust competently. But it is not. And accordingly, to the extent that we value trusting well and not just trusting successfully, we need to be able to say what trusting well involves.

Here an analogy to the value of knowledge literature is helpful: epistemologists care about more than mere true (i.e., successful) belief; unsurprisingly, the value of knowledge (roughly: a belief whose correctness is in some way through ability¹³) is widely taken to exceed the value of mere true belief. Not implausibly, there is a species of trust the value of which is greater than merely successful trust. If the value of knowledge debate is any guide,¹⁴ we’ll locate such a value by first looking at (i) what trusting well, viz., trusting competently, involves, and then to (ii) successful trust that is connected to competent trusting in the right kind of way.

5 .

A competence is, in short, a disposition to perform well in a given domain.¹⁵ What counts as performing well, or reliably enough, depends on the domain of performance.¹⁶ A baseball-hitting competence may require just that one hit the ball safely about 30% of the times one tries. Though a competence to ride a bike may demand a greater degree of reliability—viz., falling off the bike more than 50% of the time one tries (in normal conditions) betrays a lack of a competence to ride a bike.

On Sosa’s view, competences have a ‘triple-S’ constitution—seat, shape and situation—with reference to which three kinds of dispositions can be distinguished: the innermost competence (seat), the inner competence (seat + shape), and the complete competence (seat + shape + situation).

To bring this idea into sharp relief, just consider the illustrative example Sosa offers concerning one’s (complete) competence to drive a car:

With regard to one’s competence in driving, for example, we can distinguish between (a) the innermost driving competence that is seated in one’s brain, nervous system, and body, which one retains even while asleep or drunk; (b) a fuller inner competence, which requires also that one be in proper shape, that is, awake, sober, alert, and so on; and (c) complete competence or ability to drive well and safely (on a given road or in a certain area), which requires also that one be well situated, with appropriate road conditions pertaining to the surface, the lighting, etc. The complete competence is thus an SSS (or an SeShSi) competence (2017, 191–92).

The idea that one can retain one’s innermost competence to drive a car even while drunk or on slick roads comports well with the familiar idea—defended variously by Tony Honoré (1964), Anthony Kenny (1976) and Mele (2003, 447–70)—that one can retain a general ability to do something, $\phi$, even when one lacks a specific ability to $\phi$.

Consider, for example, a competent prosecutor, Akira, who is alert and in possession of the defendant’s deposition (and other relevant legal documents), standing in a courtroom. If we assume further that the evidence in Akira’s possession sufficiently implicates the defendant, there is nothing standing between Akira and a successful prosecution. In such a circumstance, as John Maier (2018, §1.3) puts it, every prerequisite for Akira’s successfully prosecuting the defendant is met. In such a case, Akira has the specific ability to prosecute the defendant. But now imagine an internally identical lawyer—Akira*—who is just like Akira in all respects, except that Akira* has been drugged and deprived of the defendant’s deposition and is locked in the courthouse basement. Akira* lacks the specific ability to prosecute the defendant—she is certain to fail if she tried—though she retains the general ability to do so¹⁷; she is, after all, very different from the hopeless Zakira, who, suppose, is also drugged and deprived of the defendant’s deposition and locked in the basement, but who (if not drugged, and equipped with all relevant documents, and situated in the courtroom in front of the judge) would, unlike Akira*, have no clue as to how to legally proceed, and so would fail to successfully prosecute the defendant.

How might we test for a general ability? What the foregoing suggests is that we should not ask whether one would succeed simpliciter, if they tried, but rather, if they would succeed if they tried while in conditions appropriate to the relevant performance-type.

Sosa’s own view offers a clear and helpful way to model this idea: according to Sosa, we test for an innermost competence with trigger-manifestation conditionals¹⁸; we ask, of the individual: would they perform reliably enough if they tried (i) in proper shape and (ii) properly situated? where what counts as ‘proper’ shape and situation is relative to the domain of performance.¹⁹ Because our competence-discerning judgments need to keep track of who would perform well in situations where (as Sosa puts it) ‘human accomplishment is prized (or otherwise of special interest),’²⁰ it is our own human interests and needs that—as we should expect—play a role in fixing the limits of proper shape and proper situation that circumscribe a given competence-type.²¹

6 .

What would it take, then, to have an (innermost) competence to specifically trust well? To a first approximation, a promising answer will take the following form:

Trust Competence (Trust-C): A competence to Trust-A successfully reliably enough, when in proper shape and properly situated.

What is it to be in proper shape and properly situated, vis-à-vis trusting; in what circumstances is trusting well of special human interest? Consider, by way of a comparison, how this question might be approached in the case of other dispositions: (a) the flammability of a match²² and (b) an archery competence. In each case, the relevant trigger-manifestation conditional corresponds with shape and situational parameters that track the conditions under which the relevant success (viz., hitting the aim) is of primary interest to us.

	Trigger-manifestation conditional	Shape	Situation
Flammability of a match	If it were struck it would likely light	Dry	Plenty of oxygen
Archery competence	If you shoot, you’d likely hit	Awake, sober, alert	Plenty of light, low winds
Trust competence	If you trust, you’d likely trust successfully	?	?

Consider first the shape relevant to Trust-C. Presumably, this will involve at least certain healthy levels cognitive functioning (e.g., the sort relevant to risk assessment) which preclude various kinds of mental incapacitation.

Moreover, it’s plausible that one is not in proper shape to trust if one is the subject of some form of ex ante manipulation or coercion. Here a distinction is needed between (i) manipulation ex ante into trusting (e.g., as when one is tricked into trusting²³); and (ii) manipulation post hoc by the trustee. One may be in proper shape to trust competently even if one’s trustee ultimately happens to betray one’s trust—viz., manipulation post hoc.²⁴

However, just as we don’t test for a driving competence by checking one’s reliability in conditions where (for instance) one is misled about the correct speed limit—for example, if pranksters swapped out a 20 mph sign for a 50 mph sign—nor, in the case of a competence to trust well, do we check whether one performs reliably enough in conditions in which (for instance) one is manipulated into trusting, as when one is the subject of an elaborate prank or deception (i.e., a surprise party). These are abnormal circumstances, not the circumstances in which we generally value the accomplishment of trusting well.

Moreover, just as we don’t test whether one is a competent truster by asking whether they trust successfully reliably enough when drugged, coerced or manipulated ex ante (i.e., in improper shape), nor do we do so in situations where normal bounds of risk, effort and skill are not present. In a bit more detail: it doesn’t count against someone’s having a competence to trust well if the truster would not trust-A successfully reliably enough in conditions where the (a) risk to the truster is excessively high and gains of betrayal are enormous²⁵; or where the level of (b) effort or (c) skill that would be required by the trustee to take care of things as entrusted is abnormally high.²⁶

The foregoing then suggests we fill out the above table in the following way. We test for a trust competence by asking: would the truster trust successfully reliably enough when in normal mental shape and not manipulated ex ante into trusting (i.e., in proper shape), and in a trust situation within normal bounds of risk, effort and skill (i.e., when properly situated).²⁷

7 .

Consider now the following case:

Gettier trust: Inspector Pazzi is attempting to catch an art thief, and to do so, he relies on a museum curator, Dr. Fell, whom Pazzi knows has always been trustworthy (unlike some of the potential experts he could have relied on), to assist him. Dr. Fell, it turns out, easily could have betrayed Pazzi in this particular situation (the art thief, it turns out, is his lover) but Fell decides ultimately by flipping a coin not to betray Pazzi.²⁸ Here Inspector Pazzi trusts competently and successfully, but that he trusts successfully isn’t because of the competence he has, it’s because Dr. Fell’s coin landed the way it did.

The above case is, no doubt, a case of lucky trust,²⁹ one where the luck in play taints our assessment of Pazzi’s trusting successfully. But the luck does not resemble (in epistemology) a completely wild guess³⁰ or (in an athletic performance) a shot performed blindly, and this is because the truster here exercises a competence to trust well. We have, to put it simply, a kind of ‘justified, successful trust’ where the source the justification (or—more generally—what accounts for why the trusting is competent) has very little to do with why the trusting is successful.³¹

What this all suggests is that even when one trusts in a way that manifests competence and also trusts successfully, something still might fall short. One’s trusting successfully might still not manifest one’s competence to trust well.³²

On Sosa’s performance normativity model, when a performance’s success issues (non-deviantly³³) from a complete competence, the performance is not only successful, and competent, but apt.³⁴ This points to there being something better than trusting (merely) successfully and competently: trusting aptly, where one’s trusting successfully manifests one’s complete Trust-C competence—something Inspector Pazzi clearly lacked.

Apt trust: S trusts aptly if, and only if, S’s successful Trust-A’ing manifests S’s complete Trust-C competence.

8 .

Is apt trust then the best kind of trust we should aspire to? Recall here the axiological analogy with belief: belief that hits its aim (truth) because of ability—viz., knowledge, according to (robust) forms of virtue epistemology³⁵—is an achievement, more valuable than true belief otherwise attained.

Apt trust, likewise, is an achievement—a success through ability. But even so, apt trust may be fragile in the following respect: (i) one trusts aptly when one’s successful Trust-A’ing manifests one’s complete Trust-C competence; (ii) that one manifests a complete Trust-C competence requires that one in fact be in proper shape and properly situated; (iii) one, properly constituted and situated, might nonetheless very easily trust when not in these conditions, where doing so would not reliably lead to successful trust.

Consider now the following case:

Mr. X: Mr. X, having read The Art of the Deal along with several books by Tony Robbins, fancies himself a charismatic dealmaker, overestimating his influence. Mr. X entrusts Mrs. Y with information I, in a situation within normal bounds of risk, effort and skill, and Mrs. Y does not betray Mr. X. Mr. X’s trust-A on this occasion may be apt—his successful trust manifests his competence to trust reliably enough in normal conditions. However, suppose that while Mr. X in trusting Mrs. Y has trusted aptly, he very easily would have trusted inaptly in those conditions. Although in entrusting Mrs. Y with information I, the risk to Mr. X is in fact not excessively high and gains of betrayal are within normal bounds, Mr. X. (with a distorted view of his charisma and influence, thanks to the Tony Robbins books) would easily have entrusted Mrs. Y with information I outside such bounds (e.g., had I been information that would have given Mrs. Y huge gains if divulged with little threat of her detection), in a situation where he would not have been a reliable enough truster.

In the above case, Mr. X’s trusting is apt, though very easily would he have trusted inaptly, outside of his range of sufficient reliability, given the distorted view he has of his competence to trust, a distorted view which precludes him from accurately gauging the risks of trusting inaptly that are present. That Mr. X trusted aptly on this occasion as opposed to inaptly doesn’t owe to any awareness of his of the threshold of his own competence to trust reliably (an awareness he lacks ex hypothesi), but rather just to good fortune.

9 .

What, exactly, is ‘missing’ in the case of Mr. X? How would this best be articulated? Here it will be helpful to introduce what Sosa calls full aptness, as this applies to performances more generally:

The fully desirable status³⁶ for performances in general is full aptness: it is aptness on the first order guided by apt awareness on the second order that the first order performance would be apt (likely enough).³⁷

In order to unpack this idea, compare Mr. X with a basketball player who shoots aptly within his threshold of sufficient reliability, but unaware of where that threshold lies, very easily would have shot from outside it. In such a case, the shot is apt—viz., it issues from the shooter’s complete competence—despite the shooter lacking a second-order ability to competently gauge the risk of inaptness, something that requires (among other things) a competent view of the shooter’s own competence.

The basketball player, then—like Mr. X.— performs aptly, but not fully aptly. There is, though, a more incisive way of putting Mr. X’s situation, one that distinguishes between two aims one has in trusting.

Consider, as Sosa (2015) says of the basketball player, that he:

… aims not just to succeed no matter how aptly [sic. but] to succeed aptly enough (through competence), while avoiding too much risk of failure. Their shots are assessed negatively when they take too much risk (2015, 85).

On the one hand, the truster (like the basketball shooter) aims not merely at success, a first-order aim, but also at apt success, a second-order aim.

Domain of endeavour	First-order aim	Second-order aim
Basketball	Make basket (i.e., successful shot)	That one’s successful shot manifests one’s complete shooting competence (i.e., apt shot)
Trust	That one’s Trust-A in another is fulfilled, viz., when another has taken care of things as we have entrusted them (i.e., successful trust)	That one’s successful Trust-A’ing manifests one’s complete Trust-C competence (i.e., apt trust)

With respect to the first-order aim of trusting successfully, there is the matter of whether that is aptly attained (i.e., whether one’s trusting successfully manifests a complete Trust-C competence). But it is a distinct question whether the further aim of trusting aptly is itself attained aptly. If so, this will not just be a matter of one’s successful trusting manifesting a Trust-C competence—viz., a competence to trust-A successfully reliably enough when in proper shape and situation; aptly hitting the aim of trusting aptly requires also manifesting a second-order or reflective competence—call this meta-Trust-C—a competence to trust not just successfully, but aptly, reliably enough. Unlike the first-order Trust-C competence, a meta-Trust-C competence will be a kind of second-order ‘monitoring’ competence,³⁸ one by which the truster reliably judges risk of inaptness, as opposed to merely trusting reliably enough when in proper shape and properly situated (viz., within normal bounds of risk, effort and skill).

With reference to the foregoing, we can now capture a richer level of trust—call it convictive trust—as follows³⁹:

Convictive trust (i.e., fully apt trust): A subject S’s trust is convictive if and only if S’s successful Trust-A’ing (i) manifests one’s complete Trust-C competence (i.e., is apt); and (ii) is guided to aptness by the truster’s (second-order) assessment of risk through meta-Trust-C.

A notion that needs unpacking, of course, is that of guidance as it features in the above account of convictive trust. Guidance is needed to close a certain gap that might otherwise be present between (i) one’s apt trusting; and (ii) one’s apt (second-order) assessment of risk of trusting inaptly. Consider the following:

Sherlock: Sherlock trusts Mrs. Hudson to complete an important task, the trust is successful (she takes care of what he entrusted her to do as he entrusted her to do it), and the trust is apt: his trusting successfully manifested a complete Trust-C competence. Suppose further that his trusting was aptly risk assessed; Sherlock aptly appreciates that not easily would he trust inaptly in these conditions. But because life has gotten a bit too boring, Sherlock decides whether to actually trust by flipping a coin, and so his apt risk-assessment in this case is in fact disconnected from his apt trusting.⁴⁰

What separates convictive trust, trust that is fully apt, from trust like Sherlock’s, which is merely apt and aptly assessed for risk of inaptness—is that it is not merely apt ‘in the light of’ an apt risk assessment (through a competent view of the subject’s own competence), but also guided by that assessment—viz., where the trusting itself is because of, and not merely compresent with, the second-order risk assessment.

10 .

Fully apt trusting, despite its being apt on two levels, depends—like any kind of human performance—on a certain kind of background. The most capable poker player, for instance, seems as though she could play her hand fully aptly—apt on both levels—even if the electricity grid responsible for the ambient lighting very easily could have gone out, ruining the hand, when it luckily happened to work normally, and when the poker player simply took for granted that it would work properly.

This raises an important question for our account of trust: what kinds of things can a truster non-negligently ignore—and simply assume to be in place—when exhibiting convictive (i.e., fully apt) trust?

This is a complicated issue, and I believe it is best approached via a performative analogy from Sosa’s Judgment and Agency. Here’s Sosa:

The athlete needs to consider various shape and situation factors: how tired he is, for example, how far from the target, and so on, for the many shape and situation factors that can affect performance. But there are many factors that he need not heed. It is no concern of an athlete as such whether an earthquake might hit, or a flash tornado, or a hydrogen bomb set off by a maniac leader of a rogue state, and so on. As an athlete, he is not negligent for ignoring such factors (2017, 191, my italics).

There is, plausibly, a distinction within the class of factors that could cause a given performance to fail, between

• (i) the kinds of things a fully apt performer must heed in order to safeguard against credit-reducing luck; and

• (ii) the kinds of things he or she is free to non-negligently assume are already in place.

Sosa refers to the kinds of things in category (ii) background conditions.⁴¹ Cobbling together from his various descriptions of them,⁴² we can identify five key features of background conditions: logical, functional, modal, epistemological and normative.

Logical Background conditions are entailed by the presence of pertinent seat, shape and situation conditions; they must hold if the relevant ‘S’ [seat/shape/situation] is in place at the time of the performance.⁴³

Functional Background conditions are orthogonal to a performance qua the kind of performance it is.⁴⁴

Modal Background conditions need not hold safely

Epistemological When a background condition holds, the performer need not know that it does.

Normative The quality of a performance is not reduced or in any way effected by reducing the safety a background condition.

The epistemological and normative features of background conditions are of special importance, as they capture the threshold of permissible ignorance in fully apt trusting—one may take for granted whatever, not related to the performance-type qua performance-type—is entailed by the relevant ‘S’ [seat/shape/situation] being place at the time of the performance.

We’ve seen what fully apt (i.e., convictive) trust demands of a truster. With reference to the above account of background conditions, let’s now investigate what such trust permits.

11 .

Let’s use the following case as a reference point:

Loan Payment: A trusts B to pay back a (modern, online) financial debt, which B repays as entrusted to do. Let’s assume further that all conditions for fully apt (convictive) trust are met. So A’s trusting B on this occasion is apt—and even more—guided to aptness by A’s assessment of risk through meta-Trust-C.

What, exactly, are the background conditions in Loan Payment, conditions which we may suppose that A could freely and non-negligently assume to be in place, and which—even if they didn’t hold safely—this wouldn’t count against the quality of A’s trusting?

As noted in the previous section, background conditions are entailed by the presence of pertinent seat, shape and situation conditions in the sense that they must hold if the relevant ‘S’ [seat/shape/situation] is in place at the time of the performance. Let’s focus on what is entailed by the presence of the relevant (a) shape and (b) situation in the case of Trust-C.

• (a) Shape: Background conditions

One is in proper shape, vis-a-vis trusting, only when at least (i) mentally fit (i.e., not physically or mentally incapacitated in a way that would have a deleterious effect on the reliability one’s trusting) and (ii) when not coerced or manipulated ex ante intro trusting (see §6). Implied by the relevant sort of mental fitness is the presence of healthily functioning neurotransmitters; suppose that, had A been slipped (for instance) Alpha-PVP (i.e., flakka), A’s neurotransmitters would be functioning abnormally and would have impeded A’s ability to understand details relevant to A’s trust exchange with B. Even if A were such that very easily A may have been (undetectably) slipped flakka prior to the trust exchange which would have ruined A’s shape, but by luck was not, A may (while trusting fully aptly) non-negligently take for granted that this was not so.

A precondition of S’s not being manipulated or otherwise coerced into trusting ex ante is that A is capable of exercising base-level autonomy with respect to whom to trust; if (for instance) A were suddenly hypnotized, A would not be in such an autonomous position—A’s capacity to avoid manipulation and coercion would be undermined. Even if A very easily could have been, prior to trusting, subjected to undetectable shape-ruining hypnosis, A may, while trusting fully aptly, non-negligently take for granted that this was not so.

• (b) Situation: Background conditions

More complex and perhaps also interesting, ethically and epistemologically, are the background conditions that underlie the situational component of the Trust-C competence: normal bounds of skill/risk/effort (see §6). Let’s take each in turn.

Skill. Thanks to working online banking services, replaying a complex modern loan as entrusted doesn’t require the trustee have specialised knowledge, e.g., of partial differential equations, stochastic calculus, etc., that are beyond normal intellectual capacities⁴⁵; doing so requires performing a sequence of simple online steps within what is usually an intuitive choice architecture design; were online banking services to suddenly fail (and experts to charge exorbitant fees), the trustee could successfully (and with suitable accuracy) pay back and properly document the complex loan only through possessing abnormally high financial and mathematical skill. Even if world online banking systems easily could have crashed beyond A’s ken but did not, A may, while trusting fully aptly, non-negligently take for granted that this was not so.

Risk Thanks to a well-functioning stock market and currency system, A would not be abnormally exposed if the debt were not paid back; if the market suddenly crashed, A would be so exposed (and likewise, suppose, B would stand to gain significantly more through betrayal than B does with the stock market failing to surprisingly crash). Even if, beyond A’s ken, a cabal of oligarchs easily could have, but just so happened to not, sabotage the stock market and world currency values in a way that would have left B standing to gain much more by betrayal than otherwise, A may, while trusting fully aptly, non-negligently take for granted that this was not so.

Effort On the supposition that A is easily locatable by B through normal means, it is does not take much effort for the trustee to repay the debt to the truster; if A’s identity had been stolen, even a mathematically skilled and well-intended trustee would have to expend considerable effort to repay A rather than A’s identity thief. Even if, beyond A’s ken, a group of hackers might easily have, but did not, steal A’s identity in a way that would have required enormous effort by B to locate A to repay the loan, A may, while trusting fully aptly, non-negligently take for granted that this was not so.⁴⁶

In sum, the fully apt truster can non-negligently take for granted background conditions that are implied by the obtaining of the relevant seat, shape and situation constituents of Trust-C. Even if the background conditions themselves hold unsafely beyond the subject’s ken, this is compatible with the subject trusting fully aptly. Incompatible with trusting fully aptly is trusting even when shape and situation conditions could easily not hold (but for reasons that are not down to the unsafety of background conditions)—viz., as in the case of Mr. X, but not in the case of Loan Payment.

12 .

If the foregoing view is right, then a bi-level approach is needed to distinguish between two species of trust, each of which is of human interest, and each of which is structured differently with respect to competence and risk.

In this section, I want to outline some additional benefits that such a view offers over more traditional univocal accounts of trust in ethics and epistemology.

• (a) Puzzles about reflection

One of the perennial problems about the nature of trust concerns the interplay between trust, risk and reflection. As Annette Baier⁴⁷ vividly expresses the idea:

Trust is a fragile plant […] which may not endure inspection of its roots, even when they were, before inspection, quite healthy.

To make this idea—viz., that reflection on (and/or monitoring of) the trust relationship imperils trust—more concrete, consider the following case, due to Wanderer and Townsend (2013):

Paranoid parent. A paranoid parent […] organises a babysitter for their child, and then proceeds to spend the evening out monitoring their babysitter’s antics remotely, via a ‘nanny-cam’. The paranoid parent is not only a lousy date, but also a lousy trustor; in performing the seemingly rational act of broadening the evidential base relevant to her judgments of trustworthiness, she is, precisely, failing to trust the babysitter (2013, 1).

The idea that actively reflecting on the trust relationship so as to minimise risk is itself at tension with genuine trust has been raised in various ways in the literature on the rationality of trust.⁴⁸ As McLeod (2015, §2) states the puzzle succinctly:

Since trust inherently involves risk, any attempt to eliminate the risk through rational reflection could eliminate the trust at the same time.

A bit more permissively, Faulkner (2018, §5) remarks that:

[…][T]oo much reflection can undermine trust.⁴⁹

The principal challenge for univocal accounts of trust is to explain what trust is, i.e., its nature, while—at the same time—making sense of how the kind of reflection or monitoring that would seem prima facie needed to improve the quality of trusting would not at the same time have the consequence of destroying it.⁵⁰

A pleasing feature of the bi-level account proposed is that—at least in one important respect—the account offers a way to bypass the puzzle.⁵¹ Second-order reflection improves the quality of trusting, and at the same time, it is compatible with genuinely trusting. Genuine trust seems, as the puzzle goes, threatened by monitoring of the trustee. But the principal role of reflection in the proposed account is not reflection on, or monitoring of, whether the trustee will betray one’s trust; its role in the account of convictive trust is one of self-regulation, where the object of the reflection, monitoring and awareness at the second order is in the main one’s own competence and the conditions of its exercise.

• (b) Internalism and the quality of trust

However one addresses the previous puzzle, it is widely thought—in particular among philosophers of trust sympathetic with internalist epistemology⁵²—that one ought to have good reasons for trusting others, in particular, in cases where there are significant stakes. A correlative of this idea is that successful trust, backed by good reasons for trusting, is better (from the point of view where good trusting is valued) than trust otherwise secured.

For example, Russell Hardin (2002, 12) maintains that the quality of one’s trusting is a function of the (continually) updated reasons one has that the trustee will be trustworthy, which might come from such things as inductive generalisations and past experiences, etc.⁵³ Having such an internal perspective that backs one’s trusting is surely valuable from the perspective where trusting well is valued. Though it cannot exhaust such value.

Here a comparison in epistemology is illustrative: why is it epistemically valuable (i.e., from the point of view where we care about the truth) for our beliefs be reasonable in light of one another?

Here’s Sosa (1997) in an early paper on reflective knowledge:

Coherence-seeking inferential reason, like retentive memory, is of epistemic value when combined with externally apt faculties of perception, because when so combined it, like retentive memory, gives us a more comprehensive grasp of the truth than we would have in its absence (1997, 421).

If externally competent faculties are working as they should, then having broad coherence in our beliefs is desirable in part because it is truth-conducive, and this is so ‘even if in a demon world […] coherence fails this test’ and would not promote true belief’ (ibid., 422).

Analogously with trust: trusting backed by a rational and internally coherent perspective is valuable, vis-à-vis trusting well, when externally competent faculties (viz., Trust-C) are working as they should; in such circumstances, such coherence is desirable because it is conducive of successful trust; and this is so even if in a world where trust were categorically betrayed,⁵⁴ such internal coherence would fail this test.

If the foregoing is right, the value of trusting well cannot be explained as internalists such as Hardin suggests, even if internalist considerations can (if combined with externally competent faculties) promote successful trust. By contrast, the bi-level account fits snugly with the above points.

In short: internal coherence is (a) an important feature of convictive trust and is essential to one’s (second-order) ability to place one’s first-level trust in perspective; and (b) is desirable (vis-à-vis our aims in trusting) because in our actual world (where our first-order trust competence really is reliable) it is conducive to successful trust. This is not meant to suggest that such an internal perspective is needed for apt trust—it is not (much, anyway)⁵⁵—but rather to affirm that beyond apt trust there is a better species of trust.⁵⁶

• (c) Enriching trust (and trustworthiness)

A separate point worth addressing about the quality of trust is due to Jones (2012, 72), and in particular, to her remarks about rich trustworthiness. Firstly, note that in paradigmatic cases of interpersonal trust, we trust someone with something (i.e., to take care of some particular things, as entrusted). This is sometimes called ‘three-place-trust’⁵⁷ so as to distinguish such trusting from two-place trust, as when one trusts another simpliciter.

While apt and convictive trust, as presented, are models of three-place-trust,⁵⁸ two-place trust is also valuable. Consider Jones’ remarks on two- and three-place trustworthiness, and why two-place trustworthiness may be of special interest.

When we talk about cultivating trustworthiness, we sometimes have three-place trustworthiness in mind, as when, for example, we talk about ways of fostering the trustworthiness of doctors with respect to their patients. However, often, when we talk about cultivating trustworthiness, our target is rich trustworthiness: we want both to increase the range of domains over which people will be competent and responsive to dependency and to improve those capacities required to have a reliable grasp of these zones of competence and to be able to signal them to others (2012, 72).

By parity of reasoning, in the case of cultivating the capacity to trust well, our target is an analogous one—viz., we value increasing (i) the range of domains over which people will trust competently; but in addition to this, we also aim ‘to improve those capacities required to have a reliable grasp of these zones of competence [sic to trust well] and to be able to signal them to others’ (ibid., 72, my italics).

It should be emphasised that if the capacity to trust well is likewise enriched by way of (i) and (ii), as it plausibly would, then this will require not merely the cultivation of Trust-C, but also the cultivation of meta-Trust-C through which one has a reliable grasp of one’s zones of (Trust-C) competence.

The bi-level account neatly explains why trust would be enriched by cultivating both Trust-C and meta-Trust-C. The cultivation of the former is conducive to apt trust, which is good. The cultivation of the latter, when apt trust is secured, is conducive to convictive trust, which is even better.

• (d) Varieties of risk assessment

We’ve already noted that trust and risk are conceptually connected⁵⁹; trusting inherently involves vulnerability to the risk that the trustee will not take care of things as entrusted to.

However, the varieties of risk assessment that befit a good truster are not limited to assessing this risk—that the trustee will not take care of things as entrusted—even under a wide construal of what kinds of factors are relevant to this particular risk. There are other risks inherent in trusting.

Following Duncan Pritchard (2015, 437) it will be helpful to distinguish between (a) a risk event⁶⁰; and (b) riskiness. A risk event (alternatively called a ‘risk’) is a possible but unwanted or harmful outcome, whereas an event is risky (or ‘high risk’) when the chance of the risk event materialising is higher than is normal for an activity of the relevant kind.

Harms internal to a domain are fixed by the aims of that domain. Trusting, on the account proposed (i.e., §9), has two aims—the first-order aim of trusting successfully, and the second-order aim of trusting aptly. The non-obtaining of these aims are distinct risk events. There are thus two kinds of trust-relevant riskiness: that which features when the chance⁶¹ of the non-obtaining of the first-order aim (successful trust) is higher than normal; and that which features when the chance of the non-obtaining of the second-order aim (apt trust) is higher than normal.

Risk management techniques appropriate to each aim may come apart. One may manage first-order risk (by limiting first-order riskiness) while lacking any good way to manage second-order risk. This is the situation we find in cases like that of Mr. X., and more generally, in cases where a truster possesses, and exercises, Trust-C in ways that are not fully apt. Conversely, in cases like Sherlock, one’s second-order risk management techniques may not bear on one’s management of first-order risk.

An advantage that the bi-level account enjoys over univocal accounts of trust is that it can explain in a straightforward way why both varieties of trust-relevant risk management, neither of which entails the other, should be valued and cultivated.

• (e) The distinction between trust and reliance

What is the difference between trust and mere reliance? One familiar line adverts to trust’s involving a dependence not only on another to take care of something, but on another to do so compresently with certain attitudes—viz., with goodwill.

As Baier (1986) remarks, trust:

[…] seems to be reliance on their good will toward one, as distinct from their dependable habits, or only on their dependably exhibited fear, anger, or other motives compatible with ill will toward one, or on motives not directed on one at all (1986, 234).

In a similar spirit, Karen Jones (1996) writes:

One can only trust things that have wills, since only things with wills can have goodwills-although having a will is to be given a generous interpretation so as to include, for example, firms and government bodies. Machinery can be relied on, but only agents, natural or artificial, can be trusted (1996, 14).

Furthermore, concerning the appropriateness of our attitudes toward misplaced trust compared to misplaced reliance, Hawley (2014) writes:

Suppose I trust you to look after a precious glass vase, yet you carelessly break it. I may feel betrayed and angry; recriminations will be in order; I may demand an apology. Suppose instead that I rely on a shelf to support the vase, yet the shelf collapses, breaking the vase. I will be disappointed, perhaps upset, but it would be inappropriate to feel betrayed by the shelf, or to demand an apology from it (2014, 2–3).

For the present purposes, I’m disputing none of these points.⁶² Rather, I want to suggest how the distinction between trust and mere reliance may in fact run deeper, and why the bi-level account is nicely situated to account for why this is so.

Consider that the satisfaction conditions for reliance are specifiable extra-agentially, in terms of whether someone, or something, does (or is) as one depends on it (him, or her) to do or to be. In the case of interpersonal reliance, we might rely on the reliably self-centred person to behave as expected; whether or not this reliance is disappointed (as opposed to betrayed) is not a matter of what the relier herself does, but of what the relied upon does.

The satisfaction conditions for trust, however, are importantly not (exclusively) extra-agential, but only partly so. Just like the basketball player who chucks it from anywhere aims, each time he chucks the ball, to make the basket, the better player aims to shoot aptly, an aim the satisfaction conditions for which include not just making it, but also making a well-selected shot, conditions satisfied in part by the agent’s own contribution to the performance. Correspondingly—and this connects with the discussion in §9—in trusting of the sort that mature humans aspire to, we aim not just at avoiding being betrayed—at successful trust however we may get it⁶³—but we also aim at apt trust, and the satisfaction conditions for the latter include not only the trustee’s taking care of things as entrusted, but the also the truster’s selecting when to trust well, conditions satisfied in part by what the trusting subject brings to the trusting.

Trust and reliance accordingly differ not only with respect to (i) the former but not the latter depending on the goodwill of the trustee; and (ii) the appropriateness of our attitudes toward misplaced trust in comparison with misplaced reliance; but also with respect to (iii) the role of the trusting subject in a specification of their respective satisfaction conditions.⁶⁴

In sum, a bi-level account of trust has much to recommend it. The view has advantages over univocal accounts of trust in matters to do with reflection, trust quality, trust-enrichment, risk-assessment and the distinction between trust and reliance.

A transition from a single level to a bi-level account of trust also has more general advantages. It allows us to see more clearly what trusting well involves, how this connects with the goals mature humans have in trusting, why these goals are worthy ones, and why it matters how we attain them.