Honesty and Bad Faith

A. Bad Faith

Here is the big idea, in broad strokes: only activities governed by certain norms can be assessed for honesty or dishonesty. One performs an activity honestly by upholding the norms that govern it; culpable violations of those norms are dishonest. Such violations are dishonest because they breach others’ entitlement to trust that you will uphold the relevant norms, given that you are engaging in an activity structured by them. Violating that trust constitutes a form of bad faith, while upholding the norms vindicates it and thereby manifests good faith. In short, honesty and dishonesty are best understood in terms of good- and bad-faith participation in these unspoken agreements.

There’s a lot to unpack here, so let’s focus on the notion of bad faith. Sartre’s discussion of bad faith is probably the most prominent in philosophy. But Sartre uses the term in a technical and narrow sense to characterize a kind of self-deception about one’s existential freedom. This doesn’t match the common usage according to which, for example, a widget-maker signs an agreement in bad faith if they know that they cannot uphold it. The widget-maker can act in bad faith in this sense even if they fully grasp their “existential freedom.” So, Sartre’s ideas will not serve us here.

Rather, what it is for someone to act in bad faith, in the colloquial sense, is for them to freely participate in an agreement while intentionally refusing to follow it. Because bad faith presupposes agreement, it offers a promising lens through which to examine the nature of dishonesty for, as we’ve just suggested, honest activities rest on unspoken agreements. When someone uses an agreement to achieve their goals without abiding by it, they exploit both the agreement and its other participants. This exploitation is what makes participation in the agreement morally objectionable in the particular manner of dishonesty – something that, as we’ve seen, cannot be secured by mere rule-breaking. Thus, when someone engages in these activities while rejecting the rules that make them possible, they act in bad faith and so act dishonestly.

Some examples will help make these abstract claims concrete. By intentionally lying, one breaks a norm that governs assertion. By breaking the norm, you fail to live up to the unspoken agreement underlying the norms that govern assertion. Moreover, because you’ve intentionally chosen to assert, you’ve entered into this unspoken agreement. But you’ve entered into the agreement with the intention to break it, by intending to lie. So your lying counts as bad faith because you’ve entered into an agreement while planning to break it.

Likewise, by intentionally cheating, you break a norm that governs the game. By breaking the norm, you fail to live up to the unspoken agreement underlying the game’s governing norms. Moreover, because you’ve intentionally chosen to play, you’ve entered into this unspoken agreement. But you’ve entered into the agreement with the intention to break it, by intending to cheat. So your cheating counts as bad faith because you’ve entered into an agreement while planning to break it.

Similarly, by intentionally stealing, you break a norm that governs property. By breaking the norm, you fail to live up to the unspoken agreement underlying property’s governing norms. Moreover, because you’ve intentionally chosen to participate in the system of property – by claiming the stolen goods as your own – you’ve entered into this unspoken agreement. But you’ve entered into the agreement with the intention to break it, by intending to steal. So your stealing counts as bad faith because you’ve entered into an agreement while planning to break it.

So understood, bad faith can, perhaps, be described as a kind of self-misrepresentation. For example, by lying you represent yourself as accepting the rules that govern assertion. But you don’t end up following those rules. Because bad faith can be described as creating a kind of false expectation, and so a kind of misrepresentation, our proposal echoes the “natural answer.” But this similarity is superficial. Properly distinguishing the proposed view from the natural answer therefore requires distinguishing two types of expectations.²²

In a descriptive sense, we sometimes expect things to behave in a certain way, akin to a future-directed belief. Evidence for these descriptive expectations is often inductive or stochastic. Their content is typically descriptive. For example, when someone expects the sun to rise tomorrow, they don’t think that the sun ought to rise tomorrow in any non-epistemic sense.

By contrast, we can also expect things – typically individuals but also institutions – to behave in a certain way in a normative sense. The content of a normative expectation is typically normative. For example, when someone expects their child to say please and thank-you in this sense, they broadly think that they ought to say please and thank-you.

We can distinguish these two kinds of expectation by their connections to different clusters of non-cognitive attitudes. For example, failures of descriptive expectations characteristically license surprise or bafflement. Failures of normative expectation, in contrast, typically license disappointment or outrage. Of course you can be both baffled and outraged at the same time, so these two forms of expectation are compatible.

The natural answer focuses on descriptive expectations. In particular, it focuses on how lying creates false descriptive expectations – this is what makes many lies deceptive. But lying doesn’t automatically falsify normative expectations: people remain entitled to expect you not to lie, in the sense that they remain entitled to believe that you ought not lie, even when you do. Therefore, instead of falsifying the normative expectation that speakers should be sincere or truthful, lying violates that expectation (de re, not necessarily de dicto but perhaps non-specific de re).²³ What characterizes dishonesty isn’t false expectation, contrary to the natural answer, but violated expectation.

To illustrate, consider that when Jim enters into an agreement with John in bad faith, Jim violates John’s normative expectation for Jim to deliver John’s coffee – that’s what makes Jim’s participation in the agreement dishonest. But Jim’s bad faith doesn’t – or needn’t – falsify any of John’s descriptive expectations. For example, John might know full well that Jim is a night owl, so that he’s unlikely to rise to the challenge of getting the coffee on time. He doesn’t project or forecast that Jim will follow through, he merely hopes that he will. In that case, Jim doesn’t give John a false descriptive expectation but still acts dishonestly by agreeing in bad faith.

These two senses of expectation correspond to two senses of misrepresentation. Jim doesn’t misrepresent himself to John in the sense of falsifying John’s descriptive expectations about how he, Jim, will behave. After all, John might know that Jim is unlikely to hold up his end of the bargain. But Jim’s breach of their initial agreement misrepresents his commitment to the agreement in the sense of violating John’s normative expectations. This kind of misrepresentation is not a species of lying or mendacity. Jim’s dishonesty lies in culpably breaking their agreement, not in making one of John’s beliefs false. Consequently, although bad faith might involve a kind of misrepresentation, it’s not the kind of misrepresentation central to the “natural answer.”

Our proposal also resembles the “natural answer” in a second respect. The latter not only holds that lying is the paradigmatic example of dishonesty but also that the fact that lying is dishonest is fundamental, for it analyzes other instances of dishonesty as forms of lying, broadly understood. The dishonesty of lying is a kind of basic posit underlying the natural answer.

While our proposal is consistent with the idea that lying is the paradigmatic example of dishonesty, it offers an analysis both of why lying is dishonest and of why lying is a paradigmatic form of dishonesty. Truth-telling is widely recognized as an if not the internal norm of assertion. Liars exploit this norm while participating in the agreement to uphold it by participating in declarative speech. This observation allows us to subsume the dishonesty of lies to a more basic and general conception of dishonesty: lying culpably violates others’ normative expectations borne of reliance on assertion’s internal standards, i.e., their expectation for you to tell the truth or their entitlement to trust you to tell it.²⁴ Consequently, lying’s dishonesty is not a brute fact, but a special case of the more general analysis of dishonesty.

Second, speech is a paradigm of social engagement. If an act is dishonest because it violates an important kind of social trust, then a type of speech that breaks that trust would naturally emerge as paradigmatic of dishonesty, given that speech is paradigmatic of social interaction more generally. Consequently, our proposal can not only explain why the natural answer seems so natural; it can also explain its motivation by ratifying the centrality of lying to the genus of dishonest action, more broadly.

Here’s our official statement of this view:

The Trust Model of Dishonesty What it is for S to φ dishonestly toward R is for S’s φ-ing to violate the trust to which R is entitled under the unspoken agreement that governs φ-ing, by culpably breaking its norms.

The Trust Model improves on the Norms Model. The institution of property relies on an agreement that one’s possessions should be one’s property, with allowances for borrowing, loans, renting, and so on. By possessing something that isn’t one’s property, one breaks this agreement. The Kantian thought highlights how thieves simultaneously claim the rights of property, and so participate in the institution of property, yet try to exempt themselves from its duties. By wrongfully possessing another’s property, the thief enters into a kind of bad faith agreement with others concerning the norms that regulate property, violating a special kind of trust to which others are entitled by the institution of property.

Likewise for cheating in a game. Games are goal-oriented activities in which participants voluntarily agree to abide by rules that artificially restrict the most efficient methods of achieving those goals.²⁵ For instance, the quickest way to get the ball in the hole in golf would be to pick it up and drop it in. However, golfers are prohibited from doing so by the rules. The existence of these rules depends on players’ expectations that others will abide by them.

The Trust Model also accounts for the challenging cases presented at the outset of the paper. When someone asserts, you’re entitled to trust them to be sincere, and bullshit is insincere speech.²⁶ When you become aware of your intention to cheat at solitaire, you are in effect giving yourself permission to cheat. Consequently, you are not entitled to trust yourself not to cheat. So, you don’t act dishonestly. This contrasts with Shameless Golf Cheat where Tiger is aware of Trump’s intention to cheat, but Tiger does not permit Trump to cheat. Consequently, Tiger is entitled to trust that Trump won’t cheat, so Trump acts with brazen dishonesty by cheating.

A second attraction of this approach is its positive analysis of honesty – it offers an appealing suggestion about what it takes to be honest, beyond not being dishonest:

The Trust Model of Honesty What it is for S to φ honestly toward R is for S’s φ-ing to vindicate the trust to which R is entitled under the unspoken agreement that governs φ-ing, by following its norms.

Famously, golfer Bobby Jones lost an important golf tournament when he logged a penalty that no one else witnessed. This was an honest action. On the Trust Model, it’s an honest action because Jones vindicates the trust of his fellow golfers by following the rules of the game when he logs his penalty stroke. This seems like exactly the right thing to say about Jones’s action.

It’s also true that Bobby Jones’ representation of his score was truthful. So, the familiar-but-flawed account also accords with this paradigmatic case of being honest (if it didn’t, it wouldn’t even be a contender). But it faces trouble from games like poker because it takes truthfulness as part of the brute, primitive analysis of honesty. Poker players often bluff, misrepresenting the strength of their cards. Do you play poker dishonestly by bluffing? Clearly not. But the familiar-but-flawed analysis predicts otherwise, since it holds that misrepresentation, “lying” suffices for dishonesty.

Proponents of the analysis are aware of this obvious challenge. But rather than abandon the analysis, they take it to motivate a seemingly minor revision. For example, the following analysis has been proposed:

The Restricted Lying Model of Dishonesty (RLM) What it is for S to φ dishonestly toward R, when in “factual environments which agents take to involve expectations about the way things actually are” is for S to φ with an intention to misrepresent or deceive.²⁷

But what licenses gerrymandering the model in this way besides the need to avoid a counterexample? Why is it that an analysis of honesty can plausibly remain silent about so-called “factual environments”? What gives rise to such environments? Preserving the model in this way is akin to preserving utilitarianism by restricting the scope of utilitarianism to exclude inconvenient implications involving trolleys or the harvesting of organs.

By contrast, the Trust Model easily explains why it’s ok to misrepresent one’s position in poker but not in golf. That’s because accurately representing your play is a norm of golf. But it is not a norm of poker. As a result, others are entitled to trust you to be truthful when playing golf but not poker.

When we participate in social practices like assertion, promising, game-playing, or property exchange, we implicitly invoke norms that make those practices possible. These norms do double duty. Not only do they help characterize the relevant practices. They also structure what we expect of each other, licensing a reciprocal trust between participants in the practice. For example, speakers trust one another to assert only what they believe; players trust one another to follow the rules; co-owners trust each other to respect property boundaries. Culpably violating these norms while still presenting oneself as a participant in the practice exploits the trust generated by those norms. This kind of bad faith is what makes such behavior dishonest – not because it distorts facts, but because it exploits the normative expectations to which others are entitled within the practice.

B Nuances

We’ve said that cheating is dishonest. But imagine that we unfreeze baseball legend Ted Williams and, once his confusion and fear of the modern world abates, we trot him onto the field and ask him to do his thing. The rules of baseball have changed since Ted’s time. So, he’s likely to unintentionally break them, violating, for example, the new pitch clock requiring pitchers to pitch the ball within a certain amount of time after stepping on the mound. But he’s not playing baseball in bad faith. So, if breaking a game’s rules is cheating and cheating is dishonest, then you can be dishonest without bad faith.

However, we don’t really think Ted is being dishonest here. We doubt that breaking a game’s rules suffices for cheating, we think only culpable rule-breaking does. Culpability here requires awareness of the rules or negligent ignorance. So, Ted isn’t being dishonest.

A similar point applies in cases of unintentional misrepresentation. Suppose that Rosa is giving a museum tour and, relying on reputable training materials that nevertheless contain undiscovered falsehoods, misattributes a particular painting to Vermeer. Unbeknownst to her, the painting was recently reattributed to a lesser-known Dutch artist. The members of the tour are misled by her testimony. Nevertheless, Rosa is not being dishonest. She believes what she says and makes the claim in good faith, trusting her training and aiming to fulfill her role as a guide. The mere fact that she is mistaken does not make her dishonest. After all, dishonesty requires that she culpably betray the trust that museumgoers place in her – either by knowing that her claim is unsupported or by being negligent in failing to update her knowledge. As with Ted Williams, the absence of culpable rule- or norm-violation means that, despite the factual error, no dishonesty occurs. In both cases, the norms characteristic of the activity are violated, but without culpability, and hence without dishonesty.

Our account implies that were Rosa’s deception culpable, she would be dishonest. For example, rather than using apparently reliable (but ultimately misleading) training materials to guide her tour, she bases her tour on claims about the Dutch Golden Age found in her favorite fictional novel about Dutch Masters, which departs from actual art history.²⁸ In that case, it would be dishonest for Rosa to present herself as giving a tour of the museum. Participants on a museum tour are entitled to trust their guide to present the facts, at least to the best of the guide’s ability. Rosa culpably violates that trust by using a novel rather than reputable training materials. So, when Rosa misattributes the painting’s artist, the misattribution here is dishonest. At a minimum, whether Rosa’s misattribution is dishonest is correlated with whether she violates the relevant party’s trust to be a good tour guide. To this, we merely add the claim that the dishonesty is the product of the violation.

The cases of dishonesty we’ve used to illustrate our account feature relative agreement between the parties about the relevant rules, e.g., Jim and John agree about the content of their arrangement and playing a game like golf or chess requires knowing the rules.²⁹ But matters are not always so clear. For example, two parties may take an activity to involve different rules. Some groups require calling each ball before sinking it when playing pool whereas some only require calling the 8-ball. Does a player of the latter group act dishonestly when they don’t call each ball when they play the other group? It seems not, provided their ignorance about which rules apply is non-culpable. Consequently, non-culpable ignorance about a practice’s norms, born of reasonable interpersonal variation in the interpretation and application of those norms, often exculpates rule-breakers. Conversely, if a rule-breaker seeks to exploit this confusion – for example, by strategically calling their shot only after the ball’s direction is clear – then they are plainly not exculpatorily ignorant of the norms. After all, they can exploit the norms only because they know that they’re in flux. So, they may be acting dishonestly. Something similar holds for practices governed by vague or indeterminate norms: one breaks the rules honestly only if one’s confusion is non-culpable.

Brazen rule-breaking, we’ve argued, is dishonest when it violates a certain kind of trust. Nevertheless, we don’t claim that all brazen rule-breaking is dishonest. For example, suppose that you, a non-golfer, join your friends for a round simply for the camaraderie. You move balls, don’t count your swings, pocket your ball when you get tired, and generally break the rules at your convenience. You’re not dishonest though you’re brazenly cheating. What’s the difference between this case and the case of Trump?

Given the setup, it’s natural to suppose your friends have given you leave to cheat, just as you do for yourself when you cheat at solitaire. Because they’ve let you cheat, they’re not entitled to trust you not to. Likewise, when Tiger doesn’t give Trump leave to cheat, Trump’s cheating is dishonest. If Tiger did, it wouldn’t be.

The case raises the important question of how and when we permit others to break the rules. For example, group projects in school are governed by norms. It’s natural to divide work roughly evenly, for members to have a say in what part they accomplish, and for members to do their part, though circumstances often force us to negotiate. When a lazy group member says, “Don’t rely on me; I’m irresponsible and disorganized!” and proceeds to contribute poorly to the project, are they being dishonest? It seems not. But the norms of collaboration seem to entitle participants to trust that others will do their part, and the lazy member’s laziness seems to violate that trust, presenting a challenge for the Trust Model.

Nevertheless, though the norms of collaboration normally license trusting others to do their part, this entitlement is defeasible. If the members continue collaborating with the lazy member, particularly if they do not persistently hold the lazy member to account, then the norms governing collaboration come under dispute. By allowing the lazy member to continue collaborating without criticizing them, the lazy member could reasonably expect to have been given leave to be lazy by the group. This ignorance exculpates their rule-breaking like how ignorance of which set of pool rules governs a game can exculpate rule-breaking, as discussed immediately above.

Conversely, if the group continues to criticize the free-riding, lazy member – insisting that they either contribute or quit – then the lazy member acts in bad faith and, in doing so, dishonestly exploits their team. This scenario arises only when the group is powerless to change its membership, as when a teacher forbids regrouping. In such cases the group is vulnerable to the free-rider’s exploitation, and that exploitation is dishonest. After all, the free-rider claims credit they have not earned. Because everything is out in the open – the free-rider’s intentional exploitation and the group’s explicit criticism – there is no deception here. But, as we have seen, not all dishonesty involves deception. In this respect the free-rider resembles a brazen thief: someone who openly takes what they do not deserve. Theft is dishonest even when it is brazen, and likewise, wrongfully appropriating the fruits of others’ labour as one’s own, even when done openly, is dishonest.

Our explanation of why a dishonest action is (pro tanto) morally objectionable is that it violates a trust to which others are entitled by certain rules associated with that action. These rules can be benign, such as the rules that govern golf or chess. They can even be beneficial, such as the rules that govern a practice of considerable social good, such as assertion. But the rules that govern a practice can also be unjust. In those cases, facts about honesty may not seem closely tied to trust based on unjust rules. For example, while physicians are told to do no harm, they also have duties of disclosure to their patients. These two duties conflict when sharing a diagnosis could be harmful, e.g., a terminal diagnosis. Prior to widespread recognition of the value of patient autonomy in the 1960s onward, physicians were known to withhold terminal diagnoses. Consequently, patients were not entitled to trust their physicians to disclose these diagnoses, given the rules that governed disclosure at that time did not require it in all cases. Withholding a terminal diagnosis did not therefore violate that trust. Nevertheless, it seems dishonest to withhold such a diagnosis.

However, whether the failure to disclose is in fact dishonest depends in part on how the terminal patient regards the rules governing medical practice and what they expect of their physician as a result. For example, if the patient accepts those rules, then it seems natural to deny that a failure to disclose is dishonest. It may not respect the patient’s autonomy (or it may, in fact, if the patient endorses the rules), but it is no more dishonest than a failure to disclose one’s hand in poker. On the other hand, if the patient rejects rules that permit physicians to withhold distressing diagnoses, then they do not participate in the local practice of medical disclosure, at least as governed by the rules of 1950s medical practice. Instead, they are entitled to trust their physician to disclose their diagnosis, grounded in a broader expectation for their physician to respect their autonomy. These two cases represent two extremes on a continuum. Surely most cases will fall in the middle, and judgments of (dis)honesty will be correspondingly vague. But in this respect our account closely tracks the facts on the ground.

We’ve advocated for the Trust Model by stressing its explanatory depth. But this explanatory depth also yields explanatory breadth. Some apparent kinds of dishonesty do not involve bad faith. Consider the following case:

Wall Street Suppose that a business is cooking the books.³⁰ The auditors are coming, and the boss knows it. So, the boss asks the accountant to lie about the cooked books, which the accountant promises to do. The accountant lies to the auditors when they show up, thereby fulfilling their promise to the boss.

Cases like this, which involve multiple relations to distinct agents, are particularly challenging for the familiar-but-flawed analysis. That’s because the analysis holds that “the virtue of honesty is the virtue of being disposed, centrally and reliably, to not intentionally distort the facts as the agent sees them.” In particular, the analysis is broadly neo-Aristotelian by presupposing that virtues are, first and foremost, dispositions.

Wall Street is particularly challenging for dispositional analyses because the analysis requires us to settle whether the accountant has the relevant disposition before we can judge whether their lie is honest. This requirement makes the approach too coarse-grained for either the lie expresses the agent’s virtue, or it doesn’t. If it does, then neo-Aristotelians face the challenge of explaining why the lie contains an element of vice, given its deception. If it doesn’t, then neo-Aristotelians face the converse challenge of explaining why the lie contains an element of virtue. This dilemma pushes us to recognize a certain awkwardness in centering character when accounting for virtue.³¹

The source of this awkwardness is the simple fact that different sets of norms can overlap on a single action. The accountant’s testimony to the auditors participates equally in both the practices of speech and promise-keeping. The lie keeps a promise, so it is honest in this respect. But it violates the norms of assertion, so it is dishonest in this respect. But these respects are clearly compatible. So the case leads us to believe that honesty and dishonesty are relational properties between actions and sets of norms.

Dispositions – at least the dispositions thought to underlie Neo-Aristotelian virtues – are not, by contrast, thought of as relational. They are naturally understood as monadic. This is why the Neo-Aristotelian approach has such trouble with the case.

By contrast, actions are especially well-suited to capturing honesty’s relational nature for actions often have multiple patients. For example, a single transaction can benefit one party but harm another. The Trust Model incorporates this feature of actions. Just a single action can benefit one person but harm another, a single action can vindicate one person’s trust while violating another’s when their trust is grounded in different sets of norms. In that circumstance, our analysis predicts it will be honest relative to the first and dishonest relative to the second, ratifying common sense about Wall Street.³²

Another instance of this difficulty arises when someone “games the system.” This happens when someone exploits a system of laws while respecting the law as written while contravening its spirit, especially when the laws allow for loopholes or are subject to interpretation. For example, suppose that a free-rider creates a chain of shell companies to exploit a tax loophole that reduces her liability to almost nothing, fully complying with the written tax code while deliberately undermining the code’s purpose of fair contribution by the wealthy. The notion of exploitation here cannot be a legal one, for the behaviour is legal, by hypothesis. Rather, what’s objectionable about the free-rider’s behaviour is that they’re expecting others to contribute taxes, so they can enjoy the public goods supported by those taxes, while contributing nothing themselves.³³

A second example helps illustrate the phenomenon.³⁴ Cricketers will sometimes walk from the crease even when they have not been given out by the umpire when they know that the ball touched the bat before it was caught. These cricketers intentionally break the rules of the game. But they are not acting dishonestly – after all, there is an advantage in waiting to see whether the umpire noticed that the ball touched the bat. Indeed, surely their rule-breaking is a means of being admirably honest! Our account may seem unable to capture either fact.

Consider the first fact: does our account entail that the batter is dishonest because they intentionally violate the rules of cricket by walking from the crease? Whether our account implies this depends on whether walking from the crease in the case violates other cricketers’ expectation for batters to stay in the crease in that circumstance. Since cricketers are seldom sanctioned, if ever, for leaving the crease in this circumstance and since walking from the crease works against the batter’s self-interest, we find it hard to see how breaking this rule constitutes a violation of the other cricketers’ trust. After all violations are, by definition, culpable and culpable wrongdoing is blameworthy. But the batter is not blameworthy. So we conclude that it’s not a violation. Since it’s not a violation, it’s not dishonest.

Walking from the crease in this circumstance inverts gaming the system. Gaming the system involves following the rules in a way that is objectionable by another set of standards. By contrast, walking from the crease involves breaking the rules in a way that is admirable by a different standard – namely, the standard of fair play. Winning because the umpire missed a call is still winning – c.f. Maradona’s “Hand of God” – but it is dissatisfying. And insofar as players count on one another to play the game well, not merely to comply with its rules, they can expect one another to act as though they are out even if the umpire has missed it. Walking from the crease vindicates that trust.