By Trevor Brandt-Sarif (Harvard University)
You face a simple dilemma. An evil villain, Stella, has you captive on her Death-Star-like space station: either you take and ingest the Red Pill she is offering you, or she will destroy Earth. This Red Pill is a special belief pill, formulated by our villain herself. Ingest it, and you will suddenly form unshakeable beliefs that our villain is (a) an incredibly cool person with (b) a glowing sense of humor. Alas, she is not in the least bit cool, and her sense of humor is pathetically underdeveloped. This pill, she hopes, will finally win her social popularity. Not the most nefarious of villains, surely—but with quite the terrifying galactic firepower.
What should you do? Well, it seems prima facie clear that you ought to take the pill. After all, the consequences of the alternative include the killing of all living things on Earth. Note that this usage of ought seems to be a distinctly moral ‘ought’.
However, it seems clear that in a purely epistemic sense, you ought not to believe that Stella is cool, or that she has a wonderful sense of humor—after all, you have loads of evidence to the contrary. Assuming that the epistemic value attached to true or well-supported beliefs is an important thing to aim for, it might seem like this ‘epistemic ought’ gives rise to a separate and independent normative standard for action, according to which your goal is to maximize your true beliefs and minimize your false beliefs. If so, it would seem that you ‘epistemically-ought’ to refuse. For if you take the pill, you will increase your false beliefs by two (a and b, mentioned above), and a and b will be replacing two true beliefs (-a and -b) you previously held. In contrast, refusing the pill might kill seven billion people, but it also leaves your epistemic faculties blameless. It would seem, in other words, that by purely epistemic lights you should refuse the pill. In the following section, I will expand on what I have in mind by epistemic ought.
Until then, briefly grant me the assumption that not only are there epistemic normative standards governing beliefs, but that at least some of them (also) normatively govern behavior.
If, then, these two species of ought that bear on behavior both exist, we seem to have an obvious tension: one standard tells you to take the pill, the other to reject it. But if this is the case, it seems like there must be more to be said. Sure, you ethically-ought to ø, and epistemically-ought to -ø; but because they disagree, it is not at all clear what it is that you ought to do. Even if there are truly distinct subspecies of ‘ought’, it seems that you still need some ‘all-things-considered ought’ (hereby ATC ought), one that you can use to guide your actual choices.
It seems like there is no good answer here. When confronted with a choice, we all face a simple question: What shall I do? We want normative philosophy to answer this question, unambiguously. However, if there is no ATC ought, then normative philosophy cannot offer you conclusive guidance on whether to accept or reject Stella’s pill. You just have to pick, and simply fall short of either your ethical or epistemic obligations.
On the other hand, this possibility should leave you unsatisfied for two reasons. First, providing guidance is a crucial standard for success in normative inquiry, so we should prefer any plausible theory that offers guidance over one, like the above, that cannot. Second, it should be quite obvious that most people in your position would take Stella’s pill and save the earth from destruction. If there is no ATC ought, one might wonder both (a) why would everyone take the pill, and (b) why does this seem like the right decision?
If, instead, there is an ATC ought, then we face the daunting challenge of giving an account of it. If the ethical ought aims toward ethical value and the epistemic ought aims toward epistemic value, what is the value toward which the ATC ought aims? Further, if there are distinct ethical and epistemic oughts, how do they combine into the ATC ought?
Neither answer, ‘there is not an ATC ought’ nor ‘there is an ATC ought’, seems to be obviously right. The former leaves something very much to be desired, while the latter leaves a potentially insurmountable challenge. This paper takes on the second horn, outlining what I will call the Equivalent Treatment View. The aim of this view will be to offer a satisfying account of the ATC ought, one that captures both our ethical and our epistemic intuitions and, at the same time, yields desirable theoretical simplicity.
2. The Voluntary Epistemic Ought
Before I begin to motivate the ATC ought, I want to first make sure that we have the idea of the epistemic ought clearly in view. To begin, reflect on the difference between an agent and one of that agent’s beliefs. This might seem obvious, but not only are they distinct entities, they are also distinct kinds of entities: the former is a person, while the latter is a certain kind of attitude, specifically the attitude of taking certain propositions to be true. Because of the difference in kind between people and attitudes, the kind of normative judgements we apply to one ought to be different than the kind of normative judgements we apply to the other. For the purposes of this thesis, I will only be focusing on epistemic judgements about agents, not beliefs.
Some epistemologists focus directly on the nature and legitimacy of normative judgements about believers. A live question is whether people have voluntary control over their beliefs, and if they do not, how that affects the way we should understand epistemic duties. This question bears on, for example, whether we can have an epistemic duty to believe some proposition p if the evidence sufficiently favors p, given that we typically do not experience voluntary control over whether to believe a proposition in the face of overwhelming evidence. Some, like Richard Feldman, believe that you can have epistemic duties even without the assumption of voluntarism—according to him, in fact, it can be the case that you ought to believe p, even if you cannot believe textitp.
However, at least some epistemic obligations must bear on behavior within our control. For if we have epistemic obligations as believers, it plausibly follows that we also have the epistemic obligation to bring about epistemic value. One seems epistemically blameworthy for actively and consciously seeking out evidence that reinforces one’s political biases, and avoiding at all costs any dissenting opinions. Likewise, one who avoids learning new information at all costs seems in the epistemic wrong. In contrast, the researcher who goes out of her way to ensure she considers only neutral data seems epistemically praiseworthy.
Similarly, we have some kind of control over how we deliberate over evidence in forming our beliefs. For example, we can purposefully slow down when we form beliefs, consciously cancel out our psychological biases, and make sure we do not fall prey to any logical fallacies. All of these choices contribute to our epistemic goal of maximizing our true beliefs and minimizing our false beliefs, and they are all within our control.We should now have in view two different kinds of epistemic oughts: those which concern epistemic states over which we have voluntary control, and those which concern epistemic states over which we do not. The Equivalent Treatment View, hereby EQTV, will aim to give an account of the voluntary epistemic oughts, because these are the only kind of epistemic obligations that inform how we ought to interact with our world. It’s only when we fall short of these kinds of epistemic obligations that we are epistemically blameworthy, for one cannot be blameworthy about something wholly beyond one’s control. Thus, for the moment we will focus solely on the oughts which dictate behavior within our voluntary control. Later on, I will explain how the EQTV understands involuntary epistemic oughts.
Finally, note that the voluntary epistemic ought seems prima facie distinct from the ethical ought. After all, this ought is derived from our epistemic judgements, and has made no mention of ethical or practical concerns. I will challenge this distinction later on in sketching out the EQTV.
3. The All-Things-Considered Ought
The desire for an ATC ought is not just the wish for the convenience of a simple answer; rather, two significant arguments motivate this pursuit. First, our study of normative philosophy is based on our experience of having choices that we need to make between options presented to us, and our desire for a normative framework with which to decide which options to pick. What we are looking for is, in fact, always the ATC ought—the ought that should decisively inform and justify our choices. Each of us faces the question, “What shall I do?” And what we want is an answer of the form “this thing—because it is the option that you ought to choose.” Perhaps the ethical and epistemic oughts are only valuable to us because, in those situations where neither is threatened by conflict with the other, they already are the ATC ought.
This last idea brings us to the second argument: the motivation that drives our desire for an ATC ought is the very same motivation that drives our desire for the ethical and epistemic oughts we already have. To see this point more clearly, it helps to consider that decision-making is sometimes quite hard, in a way that requires subtle and sophisticated normative machinery to provide us guidance. Sometimes, of course, it’s easy. In particular, tradeoffs within a domain of only similar kinds do not demand sophisticated normativity. For example, if I told you that you had to choose between spending $10,000 to cure one child’s otherwise-terminal cancer and $10,000 to cure five children’s otherwise-terminal cancers, you do not need a sophisticated system of ethics to know to choose the latter. Because the outcomes are so easily commensurable, one only needs to know if more health is desirable or less health is desirable, and then one can easily figure out the prescribed choice.
However, now imagine your choice is between spending $10,000 to cure one child’s otherwise-terminal cancer and putting that money toward building a school in a developing country without sufficient educational infrastructure. This is the kind of question that demands a more robust system of ethics, one that can help you choose between two non-similar kinds of desirable good. Any interesting moral dilemma asks the chooser to pick between options that bring about non-similar good or non-similar harm, demanding from the chooser the kind of reason or rule that helps one choose from such non-similar kinds.
This is precisely the challenge we face in situations like your confrontation with Stella. There are two non-similar kinds of value—what seems like ethical value and what seems like epistemic value—and we desire a kind of rule, or set of rules, that allow us to choose in such a context. If the motivation that drove our ethical inquiries in cases like the cure the cancer or build the school example was worth the effort there, then this very same motivation should apply here in our pursuit of the ATC ought. Furthermore, if we seem to have yielded some impressive successes in these previous inquiries, we might think it is possible to find such similar successes here as well.
4. Challenging Feldman on the ATC Ought
Richard Feldman does not believe it is possible to offer an intelligible account of the ATC ought. In section three of his paper The Ethics of Belief, he writes:
I take it that when people say things such as “Moral oughts trump epistemic oughts” they are not saying that the moral weight of epistemic oughts is less than the moral weight of other moral considerations. I believe that what they are saying is that there is some sort of generic ought that somehow encompasses moral considerations, epistemic considerations, and perhaps others, and then weighs them against one another to come up with an overall assessment. This is not any particular kind of ought. It is just plain ought. I take [some] to be suggesting that when you epistemically ought to gather more evidence and you morally ought to do something else, the moral ought “wins” and you just plain ought to do that other thing. It’s this that I just don’t understand. Of course, by this I mean to suggest that no one else understands it either. It makes no sense. (692)
The thought of his outlined in the passage above—namely, that an ATC ought makes no sense—is precisely what the EQTV will combat. However, it goes about doing this the “long way”, by sketching out a plausible theory that makes sense of exactly what he says makes no sense. I will first address the argument he makes in this section of his paper more directly.
He seems to think that the epistemic and moral oughts are fundamentally different kinds, and that therefore it is incoherent to try to make any single metric for both of them, one that we could use to adjudicate between them. However, there is a significant difference between the descriptive and the normative here that he is not sensitive to. Consider, for example, his analogy to the adjective ‘bigger’:
Suppose a child has two dolls. One is short and squat. The other is tall and thin. The child asks you which doll is “bigger”. You are faced with a problem. One doll is taller. The other is wider and, let’s assume, more voluminous. The child could be asking which is taller, which is wider, or which is more voluminous. Each of those measures could be expressed by “bigger”. We needn’t assume that there’s any meaningful measure that combines these three, yielding “bigger all things considered.” Or, more precisely, we could arbitrarily combine the various measures in any way we like. But it would be a mistake to pick any such measure as what really counts as “overall bigger.” It would make no sense to ask whether height trumps width or volume trumps height in assessing overall size. We’ve disambiguated the word “bigger”. There’s no putting the senses back together, at least no way that isn’t simply arbitrary. There’s no such thing as “just plain bigger.” There are simply the various bigger-than relations. (692-693)
He is right that, in this case, there is no clear meaning of what constitutes ‘bigger’—bigger is an adjective that, on its own, is quite vague. That said, this is because there is no clear definition of the term bigger. There does not have to be—there is no practical concern here.
But now imagine, for example, that the child explains that he is asking because his class is holding a competition: whoever brings in the biggest doll gets four homework passes. Suddenly your assumption about precision might change. You might ask, “What does your teacher count as bigger?” Perhaps he’ll say, “Oh, I think she said the tallest one.” Or maybe she meant the one that is widest, or the one with the most volume. Either way, you assume that if there is going to be a competition, there should be a precise metric against which the teacher will judge the entries.
You might think that this is not necessarily true here. Perhaps the teacher just lazily said ‘biggest’, and will just ‘go with her gut on this one.’ I offer two responses to this possibility. The first is that while it could be the case, you might think this would be a failing on her part. If she is to pick one of the many, and they are all vying for her pick, she should have a precise metric, one that is easy to judge and understandable by all involved. The second, and more significant, response here is that even if she is just going with her gut on this one, there is still some kind of metric she is using to choose one. Who knows how she combines height, width, and volume in her mental comparisons, but if she picks the one she thinks is biggest, then she must be combining them somehow. Importantly, she is not just combining them in some way, but also in only one way—otherwise, she wouldn’t be able to compare the different dolls and pick one. Sure, this might not be the most precise or reproducible metric, and it might be difficult to express how she is doing it to those around her, but she is still using a metric, and in this context, that is the metric that defines what ‘bigger’ is here. That means that when Feldman says, “There are simply the various bigger-than relations,” he might be correct for so long as we remain theoretical, but once we need to make a decision that turns on a judgement of ‘bigness’, only one of these various ‘bigger-than’ relation’ can matter.
When speaking about arbitrary ‘bigness’ in school competitions, it should not be too disturbing that the relevant metric is itself arbitrary. However, when it comes to how we ought to act, we are concerned (and reasonably so) with avoiding blameworthy actions—this seems to be one of the most important priorities we can have as human beings. Therefore, one might think there must be some way to combine the various ought subspecies, just as the teacher above needed some metric to combine height, width, and volume if she was to adjudicate in the ‘biggest’ competition. Sure, in the ‘biggest’ example, her metric might be arbitrary and unreproducible—while this is not ideal, as the teacher and judge of the competition, setting the metric like this is within her power. However, in the Stella dilemma we cannot settle for an arbitrary metric that combines the ‘oughts’, given how much higher the stakes are. 
The EQTV will allow us a non-arbitrary metric that is both necessary for practical decision-making and sensitive to the concerns that seem to inform our intuitions on separate cases. It will also bring with it significantly superior theoretical simplicity and elegance, which allows the ‘ought’ of the EQTV to be defined by reference to blameworthiness and responsibility in a way that Feldman’s idea of ought cannot. Thus, you will see that there are many more reasons to embrace the EQTV than to embrace Feldman’s skepticism.
5. Rule Consequentialism: The Resources We Need
As I stated earlier, my goal in this half of my thesis is to sketch out a plausible version of the ATC ought, and to motivate its adoption. So I must address the basic question of how to construct it at all, for it seems like the epistemic ought and the ethical ought each answer to disparate normative pressure and different fundamental values. To build a coherent ATC ought, we need to find a way to compare the two in a ‘common currency’.
This means that we would need to either (1) discover a distinct third kind of value into which we could convert both ethical and epistemic value, or (2) figure out a way to convert one of the two subspecies into the other one. I do not believe there is any productive potential in option (1), though I cannot offer a rigorous proof as to why that is. Rather, I can only think of one possible value one might use, namely prudential value, and think it will not be of any help here; I will discuss this possibility specifically when I address objections later on.5 If any other plausible values could be proposed, I would have to address them separately.
Rather, I propose that option (2) is the most promising option on which to focus. Specifically, I want to explore the possibility that we can cash out most of what we desire epistemically by drawing on resources from the ethical sphere. Later, I will argue that the EQTV gives us all of—and not merely most of—what we desire from an epistemic ought.
Recall from the introduction that I am bracketing deontology in this paper, and focusing in on consequentialist ethics. One of the prominent versions of consequentialist ethics, and the one I will rely heavily on in what follows, is rule-consequentialism. In short, it says that the right thing to do is not the act that would bring about the most good, but rather the act that comports with the rule or rules whose adoption would bring about the most good. Bringing about the most possible good is generally the core motivation of any consequentialist. Rule consequentialists are usually compelled by the idea that, because of contingent psychological biases and epistemic limitations we have as humans, everyone trying to act in accordance with what rule-consequentialism dictates would bring about more net good than everyone trying to act in accordance with what act-consequentialism would dictate.
To illustrate by example, many of us think that lying is almost always wrong, often tempting though it may be. However, act consequentialism struggles to derive this result. Act consequentialism says that a case of lying is wrong if and only if some alternative to lying would have brought about better consequences. But we can imagine many situations in which lying would bring about the best consequences. Therefore, act consequentialism permits lying far more often than our intuitions think it should. One explanation of this tension is that we think that we, qua humans, are particularly badly equipped to figure out in real time which instances of lying are the harmful ones. So rule consequentialism comes to the rescue: it allows the consequentialist to treat the practice as a cohesive set of instances, and then decide whether the practice is justified or not as a whole. If it is not, then its individual instances fall with it.
Rule-consequentialism focuses on the possible rules that everyone could follow, rules that respect and take account of our epistemic limitations and irrational biases. It then prescribes that we follow whichever set of rules brings about the best possible future world (or has the greatest chance of doing so). For example, consider the question of the rightness or wrongness of deceiving others, as mentioned above. Here are a few sample potential rules we might adopt: (a) Always lie. (b) Lie whenever it serves your interests. (c) Be honest unless lying really serves your interests. (d) Be honest or lie in each case, doing whichever seems like it’ll bring about the best result in each instance. (e) Be honest unless it seems like doing so will obviously result in a grievous harm. (f) Be honest unless etiquette says to do otherwise and there is no clear potential harm. (g) Always be honest, without any exceptions.
While (a) is certainly a rule, I assume we need not think hard about it. Similarly, though it seems like there are people who use (b) or (c) as their personal rules of thumb in regards to lying, most people would not consider these good rules to ask everyone to follow, not least because it would result in a social situation that leaves little room for trust. (d), if possible, best represents what an act consequentialist would consider justified acts of truth-telling or lying. But in practice, we are very bad at predicting the future consequences of our actions most of the time, and are very good at rationalizing what we want to do, deep down, in cases of uncertainty. So (d) fails, because it is not a rule we could reasonably expect to be able to follow. We might veer all the way to the opposite extreme then, and consider (g). Given our biases and epistemic limitations, maybe the world would be best if we just never lied. While that is a tempting move, it is blind to those rare times when we can clearly foresee a huge disparity in future consequences, those situations where being dishonest seems like it will obviously make things significantly better than being honest. Perhaps (e) and (f) better capture the kind of equilibrium we are looking for here. Sure, they both allow for discretion, and discretion opens the door to the epistemic limitations and irrational biases mentioned before. But the good caused by each disclaimer might be more significant than the harm that discretion allows in these two rules, and thus they are better rules than (g). If, therefore, (e) and (f) are in fact our best rules, a rule-consequentialist would say that we ethically-ought to follow them.
This is the power of rule-consequentialism: it allows us to treat practices as justified or unjustified. This matters for questions of justification about any practice for which some instances of it might bring about good, but more often than not bring about net bad. In the section below, I will argue that ‘wishful thinking’ perfectly fits that description.
6. Our Formulation
Let us now try this same kind of rule-consequentialist approach to the question of whether we should form beliefs in accordance with the evidence. Before we posit and compare potential rules, it is worth pausing to recognize some parallels between being honest and forming beliefs in accordance with the evidence. Being honest is conducive to producing trust and coordination among people, and true beliefs on the part of the recipient of information; in a similar vein, forming beliefs in accordance with the evidence is conducive to producing true beliefs, and preventing false beliefs, for the reasoner. Trust and coordination are important goods that help us properly navigate our social world, and are thus often necessary to bring about moral good; moreover, promoting true beliefs on the part of the recipient of information is typically more likely to produce good than its alternative. Likewise, the acquisition of true beliefs and avoidance of false beliefs on the part of the reasoner are important goods that often help one properly navigate our physical and social world, and are thus also often necessary to bring about moral good.
By this last observation, I do not only mean that you need to know who is hungry in order to feed the hungry. You also need a true belief about whether the plane is in fact safe for takeoff, or you could put a planeload of people at risk. You need a true belief about what you are allergic to, or you could put yourself into anaphylactic shock. You’d need a true belief about whether the defendant committed the crime, or you could either let a guilty criminal walk free or sentence an innocent person to prison. This is all just to illustrate that true beliefs are constantly necessary to avoid the nearly countless possible situations that would bring harm to yourself or others, should you act on bad information. Sure, not all questions of belief have direct consequences for the outcome of future events, but neither do all questions of honesty. Rule-consequentialism does not need to worry about each individual instance, but rather concerns itself with formulating manageable, implementable guidelines that will cover a large body of instances.
Having noted the many similarities shared between honesty and forming beliefs in accordance with the evidence, consider the parallel set of potential rules for the latter: (a’) Never form a belief in accordance with the evidence. (b’) Form a belief however you choose. (c’) Form a belief in accordance with the evidence, except when you do not like the resulting belief. (d’) Form a belief for each instance in whatever way will maximize future good. (e’) Form a belief in accordance with the evidence, except in cases where doing so will obviously result in a grievous harm. (f’) Form a belief in accordance with the evidence, except in cases where there is virtue in doing otherwise. (g’) Always form a belief in accordance with the evidence, without exception.
Again, (a’) and (b’) should not worry us. (c’) might be a tempting rule, but it quickly leads to more trouble than benefit; for example, one might not want to believe that one has pneumonia, but if one does not form a belief in accordance with the evidence in such a case, one is much less likely to get it treated by a doctor. (d’) might be tempting, but just like with (d) in the lying case, we are too limited epistemically to successfully figure out what belief will maximize future good in each case. This thought might lead one to (g’), which would seem tempting to an evidentialist. However, at least according to ethical rule-consequentialism, it overshoots. Because of this, (e’) and (f’) should better maximize the future good than (g’), in much the same way that (e) and (f) do a better job of maximizing the future good than (g) did back in the case of lying.
7. The Equivalent Treatment View
Just as ethical rule-consequentialism yields a defeasible obligation to be honest, so too does it yield a defeasible obligation to form beliefs in accordance with one’s evidence. If one buys into ethical rule-consequentialism, one ethically-ought to reason in accordance with the evidence, except for rare cases that fit the noted sample exceptions.
The next step is to insist that this is the only sense in which we are obligated to form beliefs in accordance with the evidence; there is not, in addition, some purely epistemic obligation to do so. Generalize that step over all ostensibly epistemic obligations, and you arrive at what I call the Equivalent Treatment View.
The Equivalent Treatment View states that epistemic obligations are simply the set of ethical obligations, under rule-consequentialism, that concern the relationship between one’s beliefs and the truth. This view does not recognize any other ‘independent’ epistemic obligations. An important consequence follows: if what seems like an epistemic obligation cannot be derived from the concerns of ethical rule-consequentialism, then the EQTV says that we simply do not have any such obligation.
For better or worse, this is not simply a method of conversion. The EQTV does not, for example, show that this fundamentally different ought, the epistemic ought, can now be understood in a way that allows systematic comparison with the ethical ought, the way one might compare two different monetary currencies by multiplying the amount you have of one of them by their exchange rate. Rather, the EQTV asserts that they are of the very same kind.
8. The Involuntary Ought
Recall that there are voluntary epistemic oughts and involuntary epistemic oughts, and the EQTV is focused solely on the former. But many epistemologists also believe there are examples of the latter: sometimes it seems true to say Person S ought to believe proposition p, even if S has no ability to believe p. For example, even after taking Stella’s pill, you still ought not to believe that she is cool. The EQTV says nothing about this kind of ought, for two reasons. The first is that they seem to offer nothing of value; the second is that they might not be real oughts at all.
To begin, the EQTV is focused on the normativity meant to inform our choices and our behavior—this is the same normativity that can render us praiseworthy or blameworthy. Our motivation for the ATC ought is our desire for decisive normative guidance when we face choices. Involuntary oughts fail to meet these criteria, and therefore the EQTV does not consider them in its construction of the ATC ought. Because involuntary oughts say nothing about the choices we ought to make, involuntary oughts have nothing interesting to offer the EQTV.
We might want to push this argument a step further—we also might doubt that involuntary oughts exist at all. Prima facie, the time-honored principle of ‘ought implies can’ (hereby OIC) would indicate that there could not be any involuntary oughts. Epistemologists committed to the involuntary epistemic ought have an obvious motivation to challenge this principle. However, realize the difficulty of their task: because their goal is to argue for the existence of involuntary epistemic oughts against the OIC challenge, they must convince us that epistemic oughts are an exception to OIC. But they cannot do that by direct reference to our intuitions about epistemic oughts, for fear of begging the question. So the most promising option is to argue that involuntary epistemic oughts are part of a broader category of ‘oughts’, and convince us that the whole category is an exception to OIC, by reference to other, non-epistemic instances of the category.
Richard Feldman tries to do just this in his paper Ethics of Belief. As he attempts to describe a category of oughts that does not conform to OIC, he settles on the possibility of ‘role oughts’:
There are oughts that result from one’s playing a certain role or having a certain position. Teachers ought to explain things clearly. Parents ought to take care of their kids. Cyclists ought to move in various ways. Incompetent teachers, incapable parents, and untrained cyclists may be unable to do what they ought to do. Similarly, I’d say, forming beliefs is something people do. That is, we form beliefs in response to our experiences in the world. Anyone engaged in this activity ought to do it right. In my view, what they ought to do is to follow their evidence (rather than wishes or fears). I suggest that epistemic oughts are of this sort—they describe the right way to play a certain role. (676)
His argument seems to unfold as follows: There is a category of oughts called ‘role oughts’. Paradigmatic examples of role oughts seem like exceptions to the OIC, indicating that the category is not constrained by OIC. There is a plausible account of why the epistemic ought falls, at least sometimes, under this category. Assuming the above is true, he has offered an account of how some epistemic oughts can be exceptions to OIC.
However, his argument is unconvincing, because when role oughts are supposedly involuntary oughts, they do not seem to be ‘oughts’ at all. It is true that we might say, ‘teachers ought to explain things clearly.’ And in the context of a teacher who can explain things clearly (hereby ‘good teacher’), it seems like that phrase expresses a normative truth. But in the context of a teacher who cannot explain things clearly (hereby ‘bad teacher’), what we mean by that is ambiguous. Perhaps all we mean is, ‘teachers ought to explain things clearly, if they can,’ so that we say nothing at all about bad teachers. We also might mean, ‘teachers ought to explain things clearly, to the best of their ability,’ which would mean that we say nothing about bad teachers, so long as they are giving their best. Neither of these would indicate that bad teachers are blameworthy.
But we could understand our statement to be conveying a kind of blameworthiness. We could be communicating a nearby thought, namely, ‘only those who can explain things clearly ought to be teachers,’ and thus be blaming bad teachers for their failure to leave the teaching profession. We also might think that some of the potential blameworthiness is not directed at the teachers; maybe ‘teachers ought to explain things well’ indicates that a principal is blameworthy for failing to fire bad teachers. Not only are all of these possibilities, but there is no reason why they cannot all be true. Perhaps this utterance expresses many thoughts at once—English can be a messy language.
That said, all of these possibilities seem more plausible than the possibility that we are saying that bad teachers are blameworthy for failing to explain things well; after all, they simply cannot do that. They lack the relevant ability, and therefore should not be blamed for failing to exhibit it. Even if you believe that teachers are obligated to explain things well, that does not mean that a bad teacher ought to be a good teacher; it means that a bad teacher is falling short of the obligations of a teacher, and therefore ought to find a more appropriate job.
Feldman’s paradigmatic examples of role oughts sound relatively reasonable. But I just offered a compelling account of these examples that did not need to violate OIC. If there are two plausible interpretations of his examples, and only one demands the theoretical sacrifice of OIC, then we ought to embrace the interpretation that leaves OIC intact. But if his non-epistemic examples are compatible with OIC, then he has not convinced us that the category of role oughts is an exception to this principle. Thus he is back to where he began, defending involuntary epistemic oughts against a compelling OIC challenge. Thus we have more than just reason to ignore involuntary oughts in our formulation of the EQTV; we have reason to doubt they exist at all.
9. Motivations and Tradeoffs
The EQTV might seem like a long-shot at first, since the epistemic ought certainly seems sui generis, and not “secretly an ethical ought”. So let me emphasize three motivations for the EQTV. In spelling them out, we will see that that we do not need to sacrifice very many of our epistemic intuitions in buying this proposal. The first motivation is familiar: this proposal offers us the most plausible account of the ATC ought we so reasonably desire. The second motivation is that we should aim to adopt the simplest theory, all else being equal. Having one species of fundamental ought is far simpler and more elegant than having multiple species of fundamental oughts.
Before I address the third and final motivation, let me pause to show how little we must give up to adopt this view. First, we must simply set aside as mistaken the prima facie intuition that the epistemic ought fundamentally differs from the ethical ought. However, given that this is precisely what is in question, it should be no surprise that this would have to be sacrificed. If there was no such prima facie resistance to this idea, the EQTV would simply not be very interesting.
In fairness, though, it seems like there might be another sacrifice—if the former sacrifice was about our intuition regarding the theoretical relationship of these oughts, the latter challenges our intuitions about case outcomes. Recall that according to the EQTV, it is not always the case that we ought to form beliefs in accordance with the evidence. Sometimes it is obvious that doing so will result in great harm, and doing otherwise will result in far better outcomes. In these cases, the EQTV at least considers it permissible to form beliefs misaligned with the evidence; it might even consider it prohibited to act ‘epistemically responsibly’ in these cases. For an evidentialist who believes that we have an epistemic obligation to form beliefs in accordance with the evidence in all cases and at all times, the EQTV falls short of delivering their desired exceptionless set of obligations.
Even so, some might reasonably think that this is a sacrifice worth making. The EQTV offers us, qua agents, the ATC ought and its decisive normative guidance for all of our choices, and also delivers maximal theoretical elegance and simplicity. Weigh that against the undesirable handful of cases in which one would not be obligated to form beliefs in accordance with the evidence, and even some evidentialists might deem this a worthy tradeoff.
However, I do not consider the presence of these exceptions to be a tradeoff at all—in fact, this consequence of the EQTV is a feature of the theory, not a flaw, and thus I consider this the third motivation, mentioned above, to adopt it. The EQTV and evidentialism differ in how they treat certain cases, namely, those cases in which believing in accordance with the evidence will cause obvious and significant harm; the EQTV does not obligate us to believe in accordance with the evidence in these cases, whereas evidentialism does. As I will illustrate in the next section, the EQTV does a better job than evidentialism at capturing our intuitions about these cases—the EQTV gets these cases right, and evidentialism does not.
In this section I will expand on this third motivation by illustrating how understanding the epistemic ought as an ethical ought maps onto several examples. Begin by recalling Stella, the evil villain who was forcing you to choose between accepting the Red Pill, which would give you false beliefs about her, or watching her destroy Earth with her superlaser. The obvious ethical ought here encourages you to take the pill. After all, selfishly refusing trivial false beliefs about how cool she is, at the cost of seven billion lives, seems like a horrible moral wrong. And yet, in so far as we have a separate epistemic duty to maximize our true beliefs and minimize our false beliefs, perhaps you have a distinct epistemic ought that is demanding you refuse the pill. If that is the case and there is no ATC ought, it seems like our normative philosophy was giving you less guidance than you require. For in facing the question—shall I take the pill?—whether you follow the dictates of the ethical ought, or rather the epistemic ought, is left open for you.
However, realistically, it should be very clear to you what it is you ought—all things considered—to do in this “dilemma”; the ethical ought, and what it is that you should obviously do, line up perfectly here. It would, in fact, be worrisome if your sense of ‘epistemic duty’ was causing you any significant amount of pause in this case. But while the facts of the case make it obvious that you should take the pill, how the facts do that would still be a theoretical puzzle without the EQTV.
One might say that the epistemic ought exists, but always yields to the ethical ought when they come in conflict. This account would, however, demand an explanation about why it is that the epistemic ought “always loses” to the ethical ought; otherwise this primacy seems quite arbitrary. What quality would the ethical ought have that would give it such lexical priority over the epistemic ought? One might say instead that it is only obvious what to do in this case because it is such an extreme example, but that there are many other cases in which it is not at all obvious. But then we wonder, how many people must be threatened with death before the epistemic ought is obviously no longer significant?
In the face of all of this mess, the EQTV answers our pill dilemma with theoretical simplicity and explanatory accuracy. It seems to zero in on the most salient features of this case, without doing any harm to the typical duty we have to form beliefs in accordance with the evidence in most everyday life.
Consider, briefly, other cases of tension between the ethical ought and the epistemic ought. For example, think of an old man diagnosed with a dangerous cancer, one who is advised by his doctor that his best shot of beating this cancer and surviving is believing, as best he can, that he can beat it. One might think there is tension between how he ethically-ought to believe earnestly that he can beat this cancer, but epistemically-ought to ask for the sober prognosis and set his expectations accordingly. However, it strikes me that there is no great mystery about what he ought to do. Believing he can beat this cancer, despite the odds, might make him fall short epistemically; this belief is certainly not well-supported by the evidence. He just ought, all things considered, to believe it anyway. Note that there is no hidden tension here. If you see the house is on fire at a fine dinner party, you ought to loudly yell for everyone to evacuate. Yelling loudly is bad according to the standards of etiquette, and thus you’d fall short of the ‘etiquette ought’—but when the house is on fire, you really ought, all things considered, to yell. Falling short of the etiquette ought is exactly what you ought to do, not merely the best of a number of bad choices. In the same way, insofar as the cancer patient falls short epistemically for believing he can beat this cancer despite the odds, he really ought to, all things considered, fall short epistemically.
Likewise, think of an underdog tennis player about to enter a high-stakes matchup. The evidence would indicate that her likelihood of winning is very low. But as anyone who participates in high-level athletics knows, really believing you can win, even despite the evidence, is necessary if one wants any hope of winning at all. On the account of the two oughts that sees them as fundamentally distinct, she ethically-ought to believe she can win, but epistemically-ought to believe she will likely lose. But if she believes that she will likely lose, then she might as well not play at all; her goal is to win, but believing that she will likely lose all but guarantees a loss before the game ever begins. In fact, realize that this does not just happen for the rare underdog; all athletes have to believe that they are going to win in order to maximize their chances of winning. Thus athletics as an entire discipline is epistemically blameworthy, by its very nature, on the account that sees the two oughts as fundamentally distinct. This strikes me as an absurd accusation to level at athletics, but it seems to be entailed by the idea that the epistemic ought is independent.
In contrast, the EQTV entails no such strange blameworthiness, on the part of the cancer patient or the underdog athlete. Again, as with our dilemma posed by the evil villain Stella, the EQTV does not prescribe blameworthiness for acting in the way that clearly comports with a rule that maximizes future good, despite that entailing, in these cases, believing counter to the evidence. This lines up naturally with many of our intuitions about these cases, far better than the alternative theory.
Here I will address two potential objections, in an effort to reduce any lingering doubt that one ought to buy into this theory.
The first challenge, mentioned offhand in a previous section, would propose an alternate way to turn the ethical and epistemic oughts into comparable like-kinds. This objection considers understanding them both as derivative oughts of the fundamental ‘prudential’ ought, the ought which encourages us to act in our own self-interest. This is, to be sure, an interesting and creative proposal, but it seems to me like it demands a different, unreasonable sacrifice. Some dilemmas seem to show that what the ethical ought demands and what the prudential ought demands are incompatible. If the ethical ought were just a subspecies of the more fundamental prudential ought, then in these dilemmas, the prudential ought would win out. However, it seems to me that when the prudential ought and the ethical ought conflict, we are always supposed to choose the ethical ought. After all, ethics plays an important role in checking our desire to act only in our own self-interest. The times when it is most important to be aware of your ethical duties and to choose to fulfill them is when your ethical duties run counter to your prudential interests. Envisioning an ethical ought that only does any work when your ethical duties are not superseded by your prudential interests is envisioning an ethical ought with very little teeth indeed.
But perhaps I am assuming a mistaken conception of prudence; for example, perhaps it is always in one’s own best interest to act in ways we typically consider ethical. Then the objection we are considering loses its teeth. For if the “prudence first” view yields exactly the same results as the EQTV, then—for the purposes of my larger project, at least—I need have no objection to it. And if it doesn’t, then that can only be because the view insists, implausibly, that prudential demands can sometimes trump ethical demands.
And to address the second objection, one might simply object that the EQTV does not place a high enough value on the truth. After all, much of what motivates the evidentialists among us is their intuition that truth is inherently valuable for us, as agents, and that wishful thinking does not properly treat the truth with the respect it deserves.
In response, I want to emphasize that our intuitions concerning what we ought to do, in a given case, are much more secure than our intuitions concerning why we ought to do that thing. In a typical (i.e., non-Stella-like) case, we can all agree than an agent has done something she ought not have if, say, she forms a belief not at all in accord with her evidence. But the important question, here, is whether the wrong-making feature of her action is that she is not properly respecting the intrinsic value of truth, or rather that she is violating the ethically-best rule for forming beliefs. Though we should trust our intuitions that she has done something wrong, we shouldn’t necessarily trust our intuitions about why it was wrong.
Above I have addressed the three most worrisome objections to my proposal. In the final section of this paper, I will attempt to draw out one last conclusion. Thus far, my proposal has demanded the resources of rule-consequentialism to construct an epistemic ought out of an ethical ought. In what follows, I will attempt to show that it is not only my view that needs consequentialism, but that consequentialism needs my view as well.
12. Trapping the Consequentialist
As explained in the section on motivations, the proposal I have sketched in the preceding pages is motivated largely by my desire for an ATC ought. Though I have assumed consequentialism thus-far, making it clear that my proposal stands or falls with consequentialism, one might think that this dependence only goes one way. That is to say, one might think that though my proposal needs consequentialism, consequentialism does not need my proposal. One might wonder what alternative the consequentialist might endorse instead. The most likely answer is that the consequentialist might say that consequentialism does not entail an ATC ought, and therefore they can reject my proposal at the very beginning, choosing instead to have two distinct subspecies of ought without any meaningful way to compare them.
In what follows, I will attempt to show two ways in which one might think this consequentialist is wrong, and why instead they need my proposal just as my proposal needs them. The first way I will try to show this will be through the lens of motivations we share in common, and the second way I will try to show this is what I will call the “reduction to actions.”
First, consider the motivations that encouraged this project from the beginning. We were troubled by the idea that we lacked the possibility of an ATC ought, and that therefore normative philosophy could not give us decisive guidance for our toughest dilemmas. This concern encouraged us to figure out if either of the two subspecies of ought could serve as the grounding of the other, and the EQTV therefore argues that the ethical ought can in fact serve as grounding for the epistemic ought.
However, realize that proponents of consequentialism focus most of their energy on trying to make it seem like a more desirable ethical framework than deontology, or dutybased ethics. One of the strongest benefits of consequentialist ethics over deontological ethics is that, at least within the ethical domain, it can always offer an ATC ought.Deontology cannot offer this because it accepts the possibility of tragic dilemmas, dilemmas in which it is impossible to fulfill all of one’s obligations and therefore in which one will always be blameworthy.12
Given that consequentialism is interested in being the more desirable of the two major ethical frameworks and that the ability to consistently yield an ATC ought is one feature in its favor, consequentialism has an interest in any proposal that can help it reinforce this advantage it enjoys over deontology. Thus it has a strong incentive to adopt my proposal.
 Feldman, Richard. “The Ethics of Belief.” Philosophy and Phenomenological Research 60, no. 3 (2000): 667-95. Web.
 Feldman, Richard. “The Ethics of Belief.” Philosophy and Phenomenological Research 60, no. 3 (2000): 667-95.
 I also take no stand on whether we have direct deliberate control over what it is we believe—I am open to the possibility. If one thinks that we do, then one should think of believing in accordance with the evidence as one of these epistemic duties.
 It is worth quickly addressing one potential objection to this characterization of moral dilemmas. One might think that the Trolley Problem, and its variants such as the Judith Jarvis Thomson Transplant Dilemma, are all about the queerness of our dilemma in the face of goods of a similar kind. However, their status as dilemmas seem to, at least often, rest of the active/passive distinction, such that the kinds are not as similar as one might wish. To illustrate with an example, the possible goods to choose between might be characterized as “allowing the trolley to kill five workers, or pulling the lever and killing one other worker instead.” This could be pressed further, and given how much has been written on the trolley problem, there is no hope of my addressing it all here. I just wanted to point out that there is a response to this surface-level objection.
 There might be additional, constitutive reasons why moral norms cannot be arbitrary—it could be argued, for example, that arbitrary norms could not function as moral norms because moral norms are often thought of as definitively universal, and without a central authority that all people agree upon and have epistemic access to, it is hard to imagine how something arbitrary could govern actions universally. I
 There are many opinions as to the precise way rule-consequentialism is supposed to work here, but I take no position on that in this paper, as a function of scope—my analysis should hold equally over any version one adopted.
 Kant’s famous “Murderer at the Door” scenario would perfectly fit the bill.
 Some might respond to this proposal by noting that the formation of our beliefs is not under our direct control, and thus this exercise is meaningless. For the moment, I would encourage such an objector to just rephrase the rule as “Acting in such a way as to maximize the likelihood that one is forming beliefs in accordance with the evidence.” As mentioned before, any ostensible epistemic duty is fair game for this process.
 For example, giving strangers the benefit of the doubt might be virtuous, even though it runs counter to strict evidentialism.
 William Clifford, for example, who was the philosopher that perhaps began the ethics of belief literature, with his paper The Ethics of Belief (1877). His famous principle, Clifford’s Principle, reads, “It is wrong always, everywhere, and for anyone to believe anything on insufficient evidence.”
 To be slightly more precise, it will always yield at least one non-blameworthy option. One might argue that if one is faced with two options and they are both equally permissible, that consequentialism has in this case failed to produce an ATC ought. I would prefer to think that consequentialism does offer such an ought, but that in this case, that ought says, “I ought to do either option.” Either way, the disagreement is uninteresting, for the EQTV offered the guidance necessary to act without risk of being blameworthy, and that is significant guidance.