From Epistemic to Moral Realism

in Journal of Moral Philosophy

If the inline PDF is not rendering correctly, you can download the PDF file here.


Both beliefs and actions can be assessed as being reasonable or unreasonable, right or wrong, good or bad. Some philosophers have thought that these commonalities are evidence for moral realism. After all, if epistemic and moral facts share all salient features, then the moral anti-realist cannot claim that moral facts are uniquely suspicious entities. If this reasoning is on point, then the best arguments against moral realism are not decisive given that we should accept epistemic facts. What it does not show, and what a realist might hope can be shown, is that if there are realistically-construed epistemic facts, then there are moral facts of the same level of robustness. That is what I hope to show here.

First, I argue that the epistemic standing of a belief depends on how diligent an inquirer the agent has been; this in turn depends on facts about the moral context of the agent’s situation. Hence moral and epistemic facts are intertwined. Some might reply that the intuitions that this argument relies upon are the product of a confusion between moral and epistemic forms of evaluation. However, we do not generally conflate moral and epistemic evaluations, so there is no good reason to think this. Second, I argue that the epistemic status of an agent’s belief depends on what evidence the agent possesses when the assessment is being made. Although evidence possession is synchronic, ethical considerations are relevant to whether or not evidence is possessed by an agent at a time. So moral facts are entangled with epistemic facts. Further reasoning will show that moral facts must have the same level of robustness as the epistemic facts that they are entangled with.

1 The Epistemic Argument

Moral realism is generally understood to be the view that (a) statements like “murder is morally wrong” purport to describe the world, (b) some of these statements are non-trivially true in virtue of normative moral facts, and (c) these facts do not depend upon the attitudes of actual or idealized human agents.1 When definitions of moral realism diverge, it is usually over the formulation of (c), though something like it is needed to exclude cultural relativism and subjectivism. Some define moral realism as including epistemic criteria, such as the claim that most humans have some moral knowledge (e.g., Cuneo 2007, 32–6). For present purposes, I will understand moral realism to be the purely ontological thesis that there are moral facts which are not constructed by human beings. The term “epistemic realism” seems a natural label to apply to the view that there are mind-independent epistemic facts – e.g., that some beliefs are unjustified (Nobis 2005, 2–11). Now consider the following argument:

Epistemic Argument

  1. If epistemic realism is true, then moral realism is true.

  2. Epistemic realism is true.

C: Therefore, moral realism is true.

Nathan Nobis (2005), Terence Cuneo (2007) and Richard Rowland (2013) advance versions of this argument, which is potentially persuasive because epistemic realism seems harder to deny than moral realism. Perhaps the denial of epistemic facts is self-defeating.2 Unfortunately, the main argument that Nobis and Cuneo provide to support the first premise, which Cuneo calls the “Parity Premise,” is not persuasive.3 Nobis, speaking of moral and epistemic judgments, writes that “an objection to understanding one kind of claim realistically is also a plausible, if not equally strong, objection to the other kind” (2005, 1–2). Similarly, Cuneo argues that if moral facts do not exist, then that must be because they would have to bear one or more of the “objectionable features” that anti-realists have identified (2007, 7–8). For example, they would have “queer” motivating powers (Mackie 1977, 38–42), would play no role in explaining our experiences (Harman 1977, 3–10), or would generate categorically normative prescriptions (Joyce 2001, 34–45; Olson 2014, ch. 6). Over the course of The Normative Web, Cuneo argues that if moral facts have these properties, then epistemic facts do as well.4

When Cuneo writes that “If moral facts do not exist, that is presumably because they would display the objectionable features” (2007, 8), he seems to mean that the failure of these anti-realist arguments entitles us to conclude that moral facts exist. Why would it not be better to suspend judgment instead? By way of analogy, imagine arguing for God’s existence on the grounds that “If God does not exist, that is presumably because of the existence of evil” and then proceeding to give a theodicy. It would be strange to make the case for theism by relying on a premise that so clearly shifts the burden of proof onto the atheist. The Nobis-Cuneo defense of the Parity Premise seems to shift the burden onto the moral anti-realist in an analogous way.

Even if the best arguments against moral facts impugn epistemic facts, it does not follow that whoever accepts epistemic facts must accept moral facts. A skeptic could say that we have defeasible reasons for rejecting whatever entities bear the objectionable features. Epistemic facts have a strike against them, but perhaps we must posit them in order to avoid self-defeat. The same is not true of moral facts. This reply weakens the anti-realist’s hand as it concedes that objectionable feature-bearing entities are possible. It also invites the response that overriding considerations of a different kind favor moral facts. Nevertheless, one can accept one kind of fact and not the other without being inconsistent. A more general dilemma is this. Do we have better reasons for thinking that epistemic facts exist than we have for thinking that moral facts exist? If the answer is no, then the moral anti-realist has nothing to fear from embracing epistemic anti-realism. If the answer is yes, then some interlocutor could reject moral facts for whatever reason they are more suspect than epistemic facts.

Perhaps we should adopt a commonsense-friendly philosophical method and treat all ordinary beliefs as innocent until proven guilty, so that a convincing rebuttal to the anti-realist is all that the realist must provide. That may be right. But an argument for the Parity Premise that does not rely on a commitment to any particular philosophical method would be superior. Realists might hope to find an argument for the Parity Premise that could potentially convince interlocutors who are not willing to concede that moral facts deserve the benefit of the doubt. In what follows, I present two arguments that show that epistemic and moral facts are entangled in such a way that the existence of some epistemic facts entails the existence of moral facts. This places the Parity Premise on firmer footing.

2 The Entanglement Argument

My first argument is as follows:

Entanglement Argument

  1. If there are epistemic facts, then those beliefs that are arrived at through diligent inquiry are in some sense epistemically better than those arrived at through negligent inquiry.

  2. In some cases, the fact that an investigation counts as diligent or negligent presupposes facts about the moral features of the agent’s situation.

C: Therefore, if there are epistemic facts, then there are moral facts.5

Although I will focus on justification, the Entanglement Argument requires only that there be some dimension of epistemic assessment that is sensitive to moral considerations. I assume that doxastic states are the most basic subjects of epistemic evaluation, but that assumption is not critical, either. In Blameworthy Belief, Nikolaj Nottelmann writes that “In a strong sense, I submit, the branch of epistemology that I here engage investigates believers, not beliefs in themselves” (2006, 34).6 The idea that epistemology directly evaluates agents is congenial to the “virtue epistemology” views of Robert C. Roberts and W. Jay Wood (2009), Linda Zagzebski (1996), and others. If that turns out to be right, then we need only modify (1) so that diligent inquiry improves the epistemic standing of agents rather than their doxastic states.

“Diligence” and “negligence” in (1) are deliberately not specified as being epistemic, moral, or generically practical evaluative terms. I am sympathetic with W.K. Clifford, who declined to distinguish between epistemic and moral senses of “ought” (1999), though I cannot collapse those two forms of evaluation at this stage without begging the question. The three things I must assume here are that (i) we can in some sense be diligent or negligent inquirers, (ii) what kind of inquirers we are is relevant to the epistemic standing of some of our doxastic states, and (iii) whether or not a decision not to seek additional information constitutes investigative negligence depends on the practical, and specifically moral, context in which the decision is being made. Readers who insist on classifying these evaluative terms may either choose to read “diligence” and “negligence” morally, in which case (2) is analytically true and (1) posits a point of ethical encroachment upon epistemic assessments, or they may read them as epistemic evaluations, in which case the opposite is true.

The idea that practical considerations encroach upon epistemic assessments, or “epistemic impurism,” has recently enjoyed significant support among philosophers (Kim 2017). David Owens (2000), Jeremy Fantl and Matthew McGrath (2002, 2009), Jason Stanley (2005) and others defend impurism on the grounds that epistemic standards depend on what is practically at stake. Stanley writes that “factors that make belief into knowledge include elements from practical rationality” hence “the distinction between practical and theoretical rationality is less clear than one might wish” (2005, 2).7 In a similar vein, Owens says that:

When we assess how sensible a belief would be, we must ask ourselves whether there is sufficient evidence to justify that belief, and it is impossible to determine how much evidence should be required to convince us of the point without considering the importance of the issue, the cognitive resources available to resolve it, and so forth. (2000, 23)

I believe that the importance of the issue being investigated is significant because of how it reflects on the investigative diligence of the agent with regard to the issue being investigated. If an issue is unimportant and inquiry is costly, then an inquirer hardly seems criticizable for declining to seek additional information. If the issue is crucial, then her unwillingness to dig deeper constitutes an investigative failure, and this failure plausibly compromises the epistemic standing of some of her doxastic states. A belief that is proportioned to the current evidence will be unjustified, or at least epistemically sub-par in some way, if the investigation that uncovered that evidence is inadequate given the circumstances.

Many philosophers accept something like this. Nicholas Wolterstorff writes that “Our obligations with respect to our believings do not pertain just to propositions that we happen to consider; often it’s our obligation to get new propositions in mind” (2010, 104).8 Hilary Kornblith writes, “When we ask whether an agent’s beliefs are justified we are asking whether he has done all he should to bring it about that he has true beliefs” (2014, 33). Bruce Reichenbach claims that “epistemic excellence” involves more than having important true beliefs and avoiding falsehoods; it also requires “engaging in appropriate truth-realizing practices” (2012, 43). Lara Buchak says that “since one can act on the world in service of epistemic or practical goals, looking for evidence seems to fall within the reaches of both epistemic and practical rationality” (2010, 108). Finally, Michael Huemer writes that “a belief might be unjustified, despite its seeming to be true, if the believer fails to exercise due caution in accepting it (fails to investigate the issue, to gather evidence)” (2001, 110). He imagines a man, creatively named Smith, who attends a lecture by a creationist and readily accepts the speaker’s conclusions without ever investigating the matter further:

…I would consider Smith’s belief insufficiently justified due to Smith’s failure to investigate the other side of the issue. In order to be justified in accepting creationism, he would also have to attend a lecture (or read a book, or some such) by an advocate of the theory of evolution, to see whether they could make evolution seem reasonable. (Ibid.)

By “insufficiently justified” Huemer means insufficient to avoid epistemic blame. To modify the case slightly, suppose that Smith attends an event where a creationist speaks, followed by a defender of Darwin’s theory. After hearing the first lecture, Smith thinks “I’m satisfied” and heads for the door, never to investigate the issue further. Grant that the creationist’s arguments are reasonable enough to provide some reason to think that creationism is true to someone who has had no other exposure to debates on the topic. Nevertheless, if Smith continues to believe that creationism is true on the basis of this evidence alone, then his belief is epistemically unjustified – or, if we follow Huemer and take epistemic justification to be scalar rather than bivalent, his belief is insufficiently justified.

Surely that is not because the belief is based on incomplete evidence, since that is true of all but the most basic logical and perceptual beliefs. Nor could the reason be that Smith knows that he lacks pertinent evidence. Recall another Smith, the man who is epistemically foiled in the Gettier case. This famous counter-example to the justified true belief account of knowledge requires that Smith be justified in believing that he will get the job that goes to his rival (Gettier 1963). Smith presumably knows that he does not know how well his rival interviewed, and he knows that this information is relevant to his prospects of getting the job. And yet we have no trouble accepting that Smith’s belief that he will get the job could be justified in a case like this. So justification does not require having access to all of the evidence that is known to be pertinent when the available evidence seems compelling.

Whereas Gettier’s Smith has a justified belief, Huemer’s Smith has a paradigmatically unjustified belief. Both men proportion their beliefs to the evidence they have, and both know that they lack pertinent evidence. Huemer suggests that Smith the creationist has evidential grounds for doubting creationism because he knows that he has not investigated both sides; presumably, he knows that investigating only one side of a disputed issue is an unreliable method of belief formation (2001, 111). But something similar could be said for Gettier’s Smith: he knows that the job is competitive and that he does not know his rival’s qualifications. We can stipulate that the probability of creationism’s being true given only the evidence presented in the creationist’s lecture, sans rebuttal, matches the probability that Smith gets the job given only the evidence of his own interview, sans knowledge of his rival. We cannot expect that either man must eliminate all grounds for doubt in order to be justified because that standard makes it nearly impossible to be justified about anything substantive. So what accounts for the difference in the epistemic standings of their beliefs?

The answer, I submit, is that only Smith the creationist is plausibly characterized as a negligent inquirer. We do not expect the Smith of the Gettier case to snoop around to see how well his rival performed before concluding that the man who will get the job will have ten coins in his pocket. His mistake is in some sense an honest mistake; the same cannot be said for the partisan who eschews disconfirming evidence or the person who cannot be bothered to investigate properly before forming an opinion. Either profile could fit Smith the creationist, and either would be grounds for evaluating his belief more harshly. As Jason Baehr writes: “What is especially epistemically good or worthy about believing in accordance with a defective or contaminated evidence base, particularly when the defects in question are attributable to one’s own cognitive defects? It would not seem much” (2011, 74). So (1), the claim that the quality of an inquiry affects the epistemic standing of beliefs, seems true.

I turn my attention to (2), the claim that moral considerations are relevant to whether an agent is a diligent or negligent inquirer. This premise gains support from the thought that we do not in all circumstances need to acquire all of the information that we could possibly acquire in order to avoid being negligent investigators.9 Consider a medical researcher who can learn something of interest only by performing gruesome, involuntary experiments on prisoners of war. If he refrains from doing this on moral grounds, I think we would not say that he chose to be an ethical person at the cost of being a negligent inquirer. We would rather say that his refusal to seek new information in this way does not render him a negligent investigator. As Wolterstorff says, “which deficiencies in a person’s system of beliefs that a person is obligated to (try to) remove or forestall are very much a matter of where that person is situated in the space of moral obligation” (2010, 105). So (2) is also true.

Returning to Smith the creationist: suppose that he leaves after the first lecture only because he receives a message informing him that his child has been injured and he is urgently needed at the hospital. As before, Smith continues to accept creationism on the basis of the evidence he has, which at this point supports creationism, until he has an opportunity to watch the rebuttal on the internet a few weeks later (and does so). At that point, Smith is surprised to find that the case for evolution is stronger than he realized and, after mulling the matter over for a while, changes his mind. Consider the epistemic standing of Smith’s belief in creationism before he has the opportunity to hear the other side. Plausibly this belief deserves a better epistemic appraisal than it would have had he no good reasons to forgo, or postpone, further investigation.

According to Nottelmann, an agent’s belief is epistemically blameworthy whenever that belief is epistemically undesirable – essentially, false or unreasonable – and the agent has no appropriate excuse (2006, 47–51). The Entanglement Argument makes no reference to a specifically epistemic form of blameworthiness, and I do not think that it requires it. Nonetheless, Nottelmann’s conceptual framework is compatible with the Entanglement Argument. In his terms, Smith’s belief in creationism may be epistemically undesirable, but his moral obligation to be at the hospital affords him an appropriate excuse for having an epistemically undesirable belief. Rik Peels examines two kinds of excuses for violations of intellectual obligations: ignorance and force (2017, 132–147). If Smith had no idea that there was any other perspective to be sought out, or if a tyrannical government made a fair investigation impossible by burning every last copy of Darwin’s Origin of Species, then we would surely not fault Smith for failing to investigate appropriately.

Nor, I think, would we label Smith a negligent investigator for declining to undertake an investigation that is an extremely costly investigation. Suppose that any inquiry into evolution would trigger horrific reprisals from the anti-evolution regime. In that case, we should either say that Smith has a legitimate excuse for failing to investigate, or that he needs no excuse because he is under no obligation to investigate in these circumstances. Either way, we would not judge him to be a negligent investigator, and this is relevant to the epistemic standing of his doxastic states. So if there is practical encroachment upon epistemic assessments, as seems plausible, then facts about the moral obligations we have are relevant to some epistemic assessments.

3 Can Impurist Intuitions be Debunked?

Epistemic purism is the idea that epistemic assessments are determined entirely by factors that increase or decrease the probability of a belief being true from the point of view of the agent (Kim 2017, 1–2). Purist views include Bayesian epistemology, the view that an agent’s beliefs should be updated in a way that is informed by Bayes’ Theorem (Bovens and Hartmann 2004), reliabilism, the view that beliefs should be formed through reliable processes or by reliable faculties (Sosa 1991, 2015; Goldman 1979), and evidentialism, the view that an agent’s degree of epistemic support for any proposition ought to correspond to the degree it is supported by the evidence at the time of the assessment (Dougherty 2014; Conee and Feldman 2009).10

Purists often reject what William Alston calls the “deontological conception” of epistemic justification (1989, 115–8), according to which epistemic norms are conceptually linked with permissibility, duty, responsibility, merit, reproach and praise. They prefer a non-deontological conception. On this view, to say that an agent is epistemically justified in believing something is simply to say that his believing it would be consistent with the norms of the epistemic domain.11 Jonathan Adler writes that “the objective version of evidentialism for full belief” is unconcerned with rationality and wisdom; rather one’s “believing that p is proper (i.e., in accord with the concept of belief) if and only if one’s evidence [objectively] establishes that p is true” (2002, 51). Trent Dougherty likewise writes that the epistemic ought “is synchronic and does not imply can and so bears no logical relation to inquiry” (2014, 154–50).12

Purists can be expected to resist the conclusions I have drawn in the previous section on the grounds that I rely on an objectionably moralized conception of epistemic justification. Sosa, who endorses a version of impurism (2015, 172–3), nonetheless observes that there is an “attractive alternative” to the impurist interpretation of “high stakes” cases discussed in the previous section:

What is affected instead by rising stakes, we might counter, is how we are willing to choose on a given basis. Thus, we might feel confident enough to judge and even to affirm publicly that the ice is solid enough to bear our weight, while still being hesitant to step on it, if the water is too cold and we fear for our lives or even just for our comfort. On this view, we are still willing to think, and even to say, that the ice is solid enough, while considering this too uncertain to justify relevant action. What is more, our judgment might even constitute ordinary, commonsense knowledge, even though this knowledge is not relevantly “actionable.” (2015, 184)

Sosa’s point is that we can insulate our epistemic evaluations from judgments about certain kinds of actions. Epistemic assessments are not affected by high stakes, but our intuitions about them might be due to a subtle conflation with ethical evaluations. A similar story could be told about cases of investigative negligence of the kind that I’ve used to motivate impurism. At the outset of “The Ethics of Belief,” Clifford imagines a ship owner who does not want to spend money refitting his ship. He manages to convince himself of his vessel’s seaworthiness by diverting his mind from contrary evidence and refusing to inquire. We know this is going to end badly. The ship owner “watched her departure with a light heart … and he got his insurance-money when she went down in midocean and told no tales” (1999, 70). Commenting on this passage, Dougherty writes, “Since [Clifford’s ship-owner] owed a duty to the passengers to protect their safety, his lack of investigation is straightforwardly wrong, morally wrong… there is nothing distinctively epistemic going on here” (2011, 153).

Dougherty is more interested in showing that there is no distinctively epistemic form of responsibility – a conclusion that I can accept – than in showing that moral facts cannot be relevant to epistemic facts. But his point can be taken as a rebuke to impurists generally: because the ship owner’s behavior is “straightforwardly wrong, morally wrong” there is no need to attribute any epistemic shortcoming to him in order to account for our negative appraisal. Dougherty believes that we should not posit any ethical encroachment into the epistemic domain without good reason because this would complicate both our moral and our epistemic theories.13 We don’t have a good reason for positing this encroachment if we can explain away impurist intuitions. In defense of the impurism, I think that it is telling that we do not generally confuse moral and epistemic evaluations. We have no trouble evaluating Eichmann’s belief that the trains will arrive at Auschwitz on time, or the terrorist’s belief that the bomb will explode when he presses the detonator, as being epistemically justified provided that each has done the necessary legwork. The intuitions that I am trying to elicit arise only when Eichmann and the terrorist fail as investigators. Such failures are far from being the most alarming moral failures that these agents exhibit, to put it mildly. Likewise, we are not tempted to say that any beliefs acquired from an investigation that is foolish or morally unjustified to undertake must be epistemically unjustified. If Smith ignores the message summoning him to the hospital to hear the pro-evolution lecture, then he does something morally wrong, but he might still learn something. It is unlikely that our epistemic intuitions are corrupted by our moral judgments only in the narrow range of cases that lend intuitive support to impurism.

Perhaps the purist will say that we should be skeptical of impurist intuitions because we have powerful independent reasons to support purism and a thinner conception of epistemic justification. The only way to refute this response would be to refute all of the arguments that have been given in favor of every kind of purism, which is beyond the scope of my project here. Instead I will focus on a particular kind of purism, evidentialism, that is often thought to be especially compelling. Scott Stapleford says that our intellectual obligation to base our beliefs on our evidence “feels almost—but not quite—like a conceptual truth” and that it is “even more obvious than the clearest moral imperatives, such as the duty to help those in need” (2013, 4065). Earl Conee and Richard Feldman, two prominent defenders of evidentialism, say that they wrote their first article in defense of evidentialism when “we noticed to our amazement that prominent epistemologists were defending views that seem contrary to evidentialism….” that they “remain mildy amazed” at the persistence of disagreement (2011, 1).

Is evidentialism really so compelling? Conee and Feldman define evidentialism “in its fundamental form” as “a supervenience thesis according to which facts about whether or not a person is justified in believing a proposition supervene on facts describing the evidence that the person has” at a time (2011, 1). Surely that is not obvious. What seems compelling is the weaker claim that believing in accordance with one’s present evidence is a necessary condition for justified belief. Only the claim that belief-evidence correspondence is sufficient for epistemic justification is in conflict with the Entanglement Argument, and we have seen that this claim is contested. I believe that similar things could be said about other forms of purism: on examination, they turn out to be less intuitive than they initially appear. Moreover, their most intuitive features are consistent with the kind of impurism presupposed by the Entanglement Argument.

If evidentialism is understood in a restricted way, as Conee and Feldman have proposed, then it is compatible with the conclusion of the Entanglement Argument, after all. Kornblith, proposing a counter-example to evidentialism, asks us to imagine that Alfred justifiably believes ‘p’ and ‘if p, then q.’ Alfred believes ‘q,’ but not because ‘q’ is validly derivable from other things that he believes. Alfred distrusts modus ponens arguments and believes ‘q’ only because he likes the sound of a sentence that expresses it (2014, 21–22; for a similar case, see DeRose (2011, 138)). Evidentialism gives Alfred a clean bill of health because his conclusion fully corresponds with his evidence, albeit accidentally. Conee and Feldman offer a conciliatory response. Evidentialism, they say, is only a theory of propositional justification (i.e., belief-evidence correspondence). There is, however, also well-foundedness, which evaluates the extent to which an agent uses the evidence she has to form the doxastic attitude in question. Alfred’s belief is propositionally justified but not well-founded and this explains our intuition that something is epistemically amiss (2011, 93).14

Making evidentialism a theory of propositional justification, not all epistemic assessments, weakens evidentialism significantly. If evidentialism is silent on some forms of epistemic assessment, then it is consistent with evidentialism to say that these other assessments are sensitive to practical and moral context, per evidentialism. The Entanglement Argument does not specify that epistemic justification is entangled with moral considerations, only that some dimension of epistemic assessment is. I have assumed that this dimension is epistemic justification, but again that assumption is not critical. Indeed, it is plausible that ethical considerations encroach upon well-foundedness. After all, the degree to which an agent has reflected diligently on the evidence she has seems important for determining the extent to which an agent uses the evidence she has to form the doxastic attitude in question. If Conee and Feldman are right, then the best evidentialist response to Kornblith’s proposed counter-example is to modify evidentialism in such a way that it is compatible with my thesis.

Although I am under no illusion that this settles the longstanding dispute between epistemic purists and impurists, I hope that I have shown that the impurist has ample resources for resisting purist attempts to debunk intuitions congenial to impurism.

4 The Evidence Possession Argument

I now turn to my second argument, which is as follows:

Evidence Possession Argument 15

  1. If there are epistemic facts, then there are facts about whether particular agents possess or lack evidence at a time.

  2. If there are facts about whether particular agents possess or lack evidence at a time, then there are moral facts.

C: Therefore, if there are epistemic facts, then there are moral facts.

Unlike the Entanglement Argument, the Evidence Possession Argument does not rely on the assumption that information an agent could have acquired, but did not due to investigative negligence, is relevant to the epistemic status of her current beliefs. (1) is intuitive enough: if there are epistemic facts, then they must include truthmakers for statements like: “The jury’s decision, although mistaken, was rational because it was based on the evidence that the jury members possessed at the time.” (2) is more contentious, and so my aim here will be to motivate it. Evidence possession is synchronic in the sense that an agent can have some evidence, and lack other evidence she could acquire, within an arbitrarily small time slice. That does not mean that evidence possession has no diachronic implications, however. For an agent to have evidence, it must be possible for her to integrate it into her thinking and decision-making, activities that take time to carry out. This leaves the door ajar for ethical encroachment.

To see how, suppose for the sake of argument that some physical objects constitute evidence. To have evidence of this kind would be to have it within one’s grasp, almost literally. The members of a jury may be said to have a murder weapon as evidence if they can request to see a piece of evidence that is inside a labeled bag in a nearby room, for example. They cannot be said to have the same object as evidence if they know it exists, but cannot retrieve it or examine it in ways that will be helpful to their purpose. Nor do I think can they be said to have it if they can only retrieve and examine it by paying an unacceptable cost, including an ethical cost. Suppose that the only way to physically bring the murder weapon in front of them and examine it is to take Smith’s family hostage and make him retrieve it at considerable risk to himself. Plausibly, that is a reason for thinking that they do not actually have the murder weapon as evidence since it can be retrieved only at an unacceptable cost.

So if a physical object can constitute evidence, as ordinary language suggests, then prudential and ethical features of a situation are relevant to whether or not one possesses some physical object as evidence at a time. Some deny that physical objects constitute evidence in the epistemically relevant sense, however. According to mentalism, “a person’s beliefs are justified only by things that are internal to the person’s mental life” (Conee and Feldman 2011, 55).16 On this view, the jury has the relevant evidence only when the physical objects classified legally as evidence cause them to have evidence-constituting mental states. Does mentalism provide a plausible route for resisting (2)?

I believe that it does not. Consider the question of which evidence-constituting mental states count as evidence that an agent has at a time. The broadest possible answer is what Feldman calls the “total view,” according to which all of the evidence-constituting mental states in a person’s mind constitute the evidence she has. Feldman recognizes that this view is subject to compelling counter-examples. If I have a memory that can only be recovered through a specific set of triggers of which I am unaware, then it seems that I do not have this memory as evidence. According to the total view, however, I do have this memory as evidence (2011, 228). Feldman rejects the total view because it has that counterintuitive implication. After considering a number of intermediate views and finding them all unsatisfactory – for good reasons, I believe – he arrives at the opposite extreme. On Feldman’s view an agent has a mental state available as evidence if and only if she is currently thinking about it (2011, 232).

Why does Feldman define “having evidence” so narrowly? Why now allow that thoughts which could be brought before the mind with minimal effort might be currently available evidence? The answer is that Feldman believes he cannot do that without compromising the independence of practical matters and the epistemic assessments of beliefs, i.e., without endorsing impurism. He asks us to consider a pair of cases. In the first, he accepts a friend’s word about the difficulty of the hike to Precarious Peak (he says it’s a breeze), failing to recall that his friend is a notorious liar and prankster. In the second case, he fails to look up information about that hike in a guidebook on hand. Feldman writes: “In each case I could have ‘looked it up’ but didn’t. Perhaps my failure to do so constitutes methodological irrationality in each case, but it doesn’t show current-state irrationality in either case” because my beliefs are in both cases based on the evidence that I currently have (2011, 234).

According to Feldman, beliefs can be said to be methodologically rational “if and only if they are formed as a result of good epistemic methods.” Current-state rationality “concerns whether believing a particular proposition is rational for a person at a time given exactly the situation the person happens to be in at the time” (2011, 228). Feldman believes that only current-state rationality is “properly epistemic” because methodological rationality is entangled with practical matters (2011, 235–6). This is demonstrated by the apparent parity of internal and external acts of investigation in his Precarious Peak examples. The upshot of treating methodological rationality as not being genuinely epistemic is that neither external nor internal acts of investigation, such as memory searching or reflection, can be relevant to the epistemic assessments of an agent’s beliefs.

It is implausible, however, that whether or not a belief was formed in accordance with the best epistemic methods has no implications for the epistemic standing of that belief on any dimension of epistemic evaluation. Indeed if epistemic assessments were entirely synchronic, then it would be hard to see how there could be such a thing as an epistemic method at all, since any method takes time to implement. Pace Feldman, methodological rationality seems to be both practically entangled and genuinely epistemic. Feldman’s narrow version of mentalism, moreover, seems no less susceptible to counter-example than the total view that he rejects. A superficially appealing, hastily-formed belief, recognizable as incorrect within a moment’s reflection, could count as justified on this view so long as the agent is not thinking of the reasons for rejecting it (even if it is clear that he should be thinking about those reasons). Feldman’s view also entails that it is logically impossible for you to fail to think about some of the evidence that you have. Intuitively, this seems not only possible, but mundane.

Feldman attempts to escape, or at least mitigate, the counterintuitive implications of his position by exploiting the ambiguities inherent in the words “currently thinking about.” He distinguishes between occasional and dispositional knowledge and claims that “currently thinking about” something is consistent with thinking about it dispositionally as well as occasionally. Feldman says that “a person knows a thing dispositionally provided that the person would know it occasionally if he thought of it” (2011, 236). This rough characterization of dispositional belief allows Feldman to say that you can be justified in believing that Washington d.c. is the capital of the United States when you are not consciously reflecting upon that fact because you are currently dispositionally thinking about it.

There are at least two problems with this proposal. First, thinking of certain things might prompt spontaneous insight, as when I immediately assent to a proposition never previously considered. Feldman seems committed to saying that such an epiphany counts as evidence possessed all along, which is absurd. Second, to say that I would know something if I thought of it could mean either that I would know it immediately upon thinking of it, or that I would know it after a period of sustained and attentive reflection. If we take Feldman to mean the former, then the boundary between possessed and unpossessed evidence appears fuzzy. The number of seconds that I need to spend thinking about a proposition before assenting to it might determine whether or not I have some mental state as evidence. If thinking of something could mean giving it sustained mental attention, then it is unclear how Feldman’s version of mentalism is substantively different from the total view that he rejects, since even deeply submerged mental states could be uncovered with enough sustained attention.

Feldman embraces a narrow notion of evidence possession because he recognizes that a more expansive view opens the door to pragmatic encroachment into evidence possession. A narrow view of evidence possession has unpalatable consequences, however, and Feldman’s strategy for avoiding these consequences is unpromising. I run Feldman’s reasoning the other way: because we cannot accept an extremely narrow view of evidence possession, we must accept pragmatic (and, I believe, ultimately ethical) encroachment into evidence possession.17 Because epistemic facts depend on facts about evidence possession, epistemic facts are entangled with moral facts.

5 What Sort of Moral Facts?

Both the Entanglement Argument and the Evidence Possession Argument purport to establish that if there are epistemic facts, then there are ethical facts of some sort. But recall that the Parity Premise of the Epistemic Argument is “If epistemic realism is true, then moral realism is true.” This goes beyond what has been established so far. The claim “Smith has an obligation to pay Jones fifty dollars” presupposes certain facts about currency, but we cannot infer from this that if we are moral realists, then we should be realists about currency, too. What reason, then, do we have for thinking that the moral facts entangled with realistically-construed epistemic facts must themselves be realistically construed? To complete the case for the Parity Premise, I offer the following argument:

Robustness Argument

  1. If moral facts are less robustly real than epistemic facts they are entangled with, then these epistemic facts will be hostage to desire.18

  2. Epistemic facts are not hostage to desire.

C: Therefore, entangled moral facts are not less robustly real than the epistemic facts that they are entangled with.

Let us return to Smith the creationist. Suppose, as you might have already suspected, that Smith has a strong desire not to investigate human origins for fear of what the inquiry might reveal, and that he breaks off his investigation for this reason. If a crude version of ethical subjectivism is true, and if moral and epistemic facts are intertwined in the way that I have argued that they are, then Smith’s desire not to believe evolution would mean that he has a moral obligation not to look into the matter. I have argued that an agent’s refusal to violate moral principles in order to undertake an investigation does not render her a negligent investigator. If that is right, then Smith would not count as a negligent investigator if he refuses to investigate for this reason. His belief in creationism could continue to be justified – an unacceptable result, surely. An agent’s belief cannot be epistemically justified simply because he desires to believe it. That would license all manner of self-deception. Hence, epistemic facts cannot be hostage to desire as they would be if ethical subjectivism were true.

What about other forms of anti-realism? The situation would be no better if epistemic facts were hostage to the collective desires of some society, as would be the case if cultural relativism turned out to be the true moral theory. The same I think would be true if fictionalism or non-cognitivism were true: realistically-construed epistemic facts would end up depending on fiction, or expressions of subjective emotion. The dependency on subjective attitudes is no less present in idealized response-dependent ethical theories.19 Matthew Kramer notes that if human beings were less compassionate, then the idealized versions of human beings would also be less compassionate (2009, 59–60). Indeed, their perfect consistency would likely make them even more malign. Were this the case, then it would follow that we would have good reasons not to undertake investigations that risk turning up information repugnant to that set of values. So on this anti-realist view, too, epistemic facts would end up hostage to contingent facts about human pro-attitudes.

One exception to this line of reasoning is moral error theory, or nihilism, the view that no non-trivial moral statements are true. On this view, realistically-construed epistemic facts would not be dependent upon individually or socially constructed moral facts, because there would be no such facts. So it seems that we are in need of a different argument against nihilism. If the Entanglement Argument is persuasive, then nihilism could be ruled out on the grounds that it generates implausible results in epistemology. That is, it would generate the result that there is no epistemic difference in Smith’s belief in the emergency case and in the case in which he has no good reason not to continue inquiring, since there are no moral facts to make the cases different. I think that is implausible, but the thoroughgoing moral error theorist must already be willing to bite far worse bullets than this (e.g., the trans-Atlantic slave trade was not morally wrong). So it is not clear that this reductio ad absurdum makes matters substantially worse for the nihilist.

To conclude, I have argued that epistemic and moral considerations are intertwined in such a way that it does not make sense to be a realist about the former and not the latter. I have also argued that, given this entanglement, the truth of most forms of moral anti-realism would make epistemic facts dependent upon subjective mental states in unacceptable ways. If what I have argued is correct, then we have good reason to accept the Parity Premise: If epistemic realism is true, then moral realism is true. Thus the moral anti-realist can reject moral realism only by rejecting epistemic realism along with it. I think we should be realists about both domains.

Biographical Note

Spencer Case has a philosophy doctorate from the University of Colorado Boulder, where he is currently a lecturer. He can be contacted at

I would like to thank the following people for helping me with previous drafts of this paper: Graham Oddie, Alastair Norcross, Chris Heathwood, Michael Huemer, Tyler Paytas, Nathan Nobis, Terence Cuneo, Stephen Kershar, Andrew I. Cohen and Jim Skidmore. The anonymous referees for this journal were also very helpful. All of these people deserve a giant heap of good karma. Apologies to anyone I am forgetting. Good karma for you, too.


For examples of philosophers who use definitions of moral realism similar to the one I use here, see Huemer (2005: 4–5), Brink (1989: 5–7), Shafer-Landau (2003: 2), and David Enoch (2013: 1–8, especially 3).

See Cuneo (2007, 117), David Copp (1995, 46–7), Bart Streumer (2013, 197), Joyce (2001, 49–50) and Stan Husi (2013, 429).

Cuneo’s statement of the Parity Premise differs slightly from mine: “If there are no moral facts, then there are no epistemic facts” (2007, 6). Rowland (2013) doesn’t offer an argument in favor of the Parity Premise but defends it against an objection by Chris Heathwood (2009).

An important caveat: Cuneo and I are operating with slightly different conceptions of “realism.” Cuneo thinks that moral realism requires that moral facts be irreducible (2007, 29–32), whereas I think that reductionism may be compatible with moral realism. Cuneo also argues that moral facts need not be mind-independent in order to satisfy “realism of a paradigmatic sort” (2007, 45–49). Presumably, the same differences would apply to his understanding of “epistemic realism,” but that is a term that he does not use. I think that these differences make no difference to my criticism of Cuneo’s argument for the parity premise or anything else that follows.

For simplification, I omitted two steps from the formalized argument: “Therefore, if there are facts about whether an inquiry counts as diligent or negligent, then there are moral facts” and “Therefore, if beliefs arrived at diligently are epistemically better than those arrived at negligently, then there are moral facts.”

See also Lorraine Code (1987, 1–14).

According to Stanley’s “Interest-Relative Invariantism,” knowledge is semantically invariantist and epistemically impurist. This means that while “knowledge” consistently refers to the same cognitive state, practical considerations are relevant to whether an agent is in that state. This contrasts with Keith DeRose’s (2011, 2009) version of contextualism, according to which practical considerations determine which doxastic state the word “knowledge” refers to, but do not change the epistemic facts themselves.

The context makes it clear that Wolterstorff doesn’t think these obligations are of radically different types.

Though a few defend extremely demanding intellectual obligations, e.g., Roderick Chisholm (1977, 14), Richard J. Hall and Charles R. Johnson (1998, 133).

Kim (2017, 2–3) classifies contextualism of the kind defended by DeRose (2011, 2009), according to which the meaning of the word “knowledge” is affected by practical factors, to be a form of impurism. But my interest here is in metaphysical impurism – how moral facts affect epistemic facts – not semantics.

I say “often” because I think that it’s possible for an evidentialist to accept the deontological conception. We can distinguish between evaluative evidentialism which simply says that a belief is good to the degree that it matches the evidence, and normative evidentialism which says that there is a duty upon the agent to bring his creedal state into harmony with his evidence. Scott Stapleford entertains, and seems to endorse, this version of evidentialism (2013, 4069–37; 2015, 84–5). If there is a positive epistemic duty for me to bring my beliefs into accord with my evidence, then this duty is plausibly subject to certain pragmatic constraints rather than being absolute and unyielding.

Evidentialism could be understood in a way that is consistent with the idea that epistemic “ought” implies can. Perhaps an agent’s inability to adjust his beliefs appropriately upon receiving new information entails that this information cannot constitute evidence for him.

Dougherty writes that once “one goes beyond actual evidence and suggests that one epistemically ought to believe on the basis of possible evidence, one will have very hard time indeed drawing a nonarbitrary line” (2014, 158). He’s right about that – it is very hard, on my view, to know how much effort one must invest in uncovering new information in order to be epistemically justified, and it will vary with context. But that is not an objection to my view; that is life.

Dougherty (2011, 155–6) uses this same reply to counter DeRose’s (2011, 138) example, which is similar to Kornblith’s. He uses the term “doxastic justification” to mean well-foundedness.

I will use the term “possessing” evidence synonymously with “having” evidence and “having available” evidence.

Mentalism canvases a wide variety of views. For David Lewis “perceptual experiences and apparent memories” constitute the evidence that I have (1996, 424). Donald Davidson likewise says that “nothing can count as a reason for holding a belief except another belief” (2001, 141). Quine identifies “all the evidence that agent has to go on” with the “stimulation of his sense receptors” (1969, 75). Timothy Williamson (1997, 717–742) defends the view that all of the propositions that an agent knows constitute her evidence. Ram Neta’s view (2008) is similar to Williamson’s. Neta agrees with Williamson that only propositions can be evidence. He adds that for an agent to have propositions as evidence she must be able to use those propositions to rationally regulate her attitudes. Moreover, rationality must require that she “distribute her confidence over hypotheses in proportion to the degree to which those hypotheses are supported by the conjunction” of all the propositions that constitute her evidence (2008, 12).

Feldman could avoid some of the counter-intuitive implications of his view of evidence possession by allowing that unpossessed evidence can affect the epistemic statuses of beliefs. If that were the case, then there would be no basis for saying that ethical considerations encroach upon evidence possession, and the Evidence Possession Argument would be doomed. However, as Feldman himself points out, pragmatic factors would then be relevant to which unpossessed evidence is relevant to justification. So we’d have epistemic-moral entanglement elsewhere. I’d like to thank an anonymous referee for making this observation.

Recall that I take moral realism to be the purely ontological thesis that there are robust moral facts. So skepticism about what the moral facts are is compatible with the kind of realism that I defend here. Expressivism and other non-cognitivist views, according to which moral statements lack propositional content, are relevant only inasmuch as they obviate the need to posit moral facts.

Cuneo allows that some forms of constructivism are compatible with realism; see The Normative Web, ch. 1.

If the inline PDF is not rendering correctly, you can download the PDF file here.

From Epistemic to Moral Realism

in Journal of Moral Philosophy



Adler Jonathan E. Belief’s Own Ethics. (Cambridge, Mass: mit Press2002).

Alston William P. Epistemic Justification: Essays in the Theory of Knowledge. (Ithica; London: Cornell up1989).

Baehr Jason . The Inquiring Mind: On Intellectual Virtues and Virtue Epistemology. (New York: Oxford up 2011).

Bovens Luc and Hartmann Stephen . Bayesian Epistemology. (New York: Oxford up2004).

Brink David . Moral Realism and the Foundations of Ethics. (Cambridge; New York: Cambridge up1989).

Buchak Lara . ‘Instrumental Rationality, Epistemic Rationality, and Evidence-GatheringPhilosophical Perspectives 26 (2010) pp. 85120.

Chisholm Roderick M. Theory of Knowledge 2nd ed. (Englewood Cliffs, nj: Prentice-Hall, Inc.1977).

Clifford W.K. The Ethics of Belief’ in Madigan T. ed. The Ethics of Belief and Other Essays (Amherstma: Prometheus1999) pp. 7096.

Code Lorraine . Epistemic Responsibility. (Hanover and London: University Press of New England1987).

Conee Earl and Feldman Richard . “Evidentialism” in their Evidentialism: Essays in Epistemology. (Oxford: Oxford up 2011) pp. 83–108.

Copp David . Morality Normativity and Society. (New York: Oxford up1995).

Cuneo Terence . The Normative Web. (New York: Oxford up2007).

Davidson Donald . ‘A Coherence Theory of Truth and Knowledge’ in his Subjective Intersubjective Objective (The Philosophical Essays of Donald Davidson) (Oxford: Clarendon2001) pp. 137–58.

DeRose Keith . ‘Questioning Evidentialism’ in Trent Dougherty ed. Evidentialism and its Discontents. (Oxford: Oxford up2011) pp. 13746.

DeRose Keith . The Case for Contextualism: Knowledge Skepticism and Context Vol 1. (New York: Oxford up2011).

Dougherty Trent . ‘The ‘Ethics of Belief’ is Ethics (Period): Reassigning Responsiblism’ in Jonathan Matheson and Rico Vitz ed. The Ethics of Belief. (New York: Oxford UP 2014) pp. 146–68.

Enoch David . Taking Morality Seriously: A Defense of Robust Realism. (New York; Oxford: Oxford up2011).

Fantl Jeremy and McGrath Matthew . ‘Evidence, Pragmatics, and JustificationPhilosophical Review 111 (2002) pp. 6794.

Fantl Jeremy and McGrath Matthew . Knowledge in an Uncertain World. (New York: Oxford up2009).

Feldman Richard . ‘Having Evidence’ in Conee E. and R. Feldman Evidentialism: Essays in Epistemology. (Oxford: Oxford up2011) pp. 83–108.

Gettier Edmund . ‘Is Justified True Belief KnowledgeAnalysis 23 (1963) pp. 1213.

Goldman Alvin . ‘What Is Justified Belief?’ in Pappas G. ed. Justification and Knowledge (DordrechtThe Netherlands: D. Reidel1979) pp. 123.

Hall Richard J. and Johnson Charles R. . ‘The Epistemic Duty to Seek More EvidenceAmerican Philosophical Quarterly 35 (1998) pp. 129139.

Harman Gilbert . The Nature of Morality. (Oxford: Oxford up1977).

Heathwood Chris . “Moral and Epistemic Open Question ArgumentsPhilosophical Books 50 (2009) pp. 8398.

Huemer Michael . Skepticism and the Veil of Perception. (Lanham, md: Rowman and Littlefield2001).

Husi Stan . ‘Why Reasons Skepticism Is not Self-DefeatingEuropean Journal of Philosophy 21 (2013) pp. 42449.

Joyce Richard . The Myth of Morality. (New York: Cambridge up2001).

Kim Brian . ‘Pragmatic Encroachment in EpistemologyPhilosophy Compass 12 (2017) pp. 114.

Kornblith Hilary . ‘Justified Belief and Epistemically Responsible Action’ in his A Naturalistic Epistemology: Selected Papers. (Oxford: Oxford up 2014) pp. 3245.

Kramer Matthew . Moral Realism as a Moral Doctrine. (Chichester u.k.; Malden, ma: Wiley-Blackwell2009).

Lewis David . ‘Elusive KnowledgeAustralasian Journal of Philosophy 74 (1996) pp. 54967.

Mackie J.L. Ethics: Inventing Right and Wrong. (Harmondsworth, ny: Pelican Books1977).

Neta Ram . ‘What Evidence Do You Have?’ British Journal of Philosophy of Science 59 (2008) pp. 89119.

Nobis Nathan . Truth in Ethics and Epistemology: A Defense of Normative Realism. (Dissertation University of Rochester 2005).

Nottelmann Nikolaj . Blameworthy Belief: A Study in Epistemic Deontologism. (Dordrecht, The Netherlands: Springer2006).

Olson Jonas . 2014. Moral Error Theory: History Critique Defense. (Oxford: Oxford up 2014).

Owens David . Reason Without Freedom. (New York: Routledge2000).

Peels Rik . Responsible Belief. (New York: Oxford up2017).

Quine W.V. “Epistemology Naturalized” in his Ontological Relativity and Other Essays. (New York: Columbia up1969) pp. 6990.

Reichenbach Bruce R. Epistemic Obligations: Truth Individualism and the Limits of Belief. (Waco, Texas: Baylor up2012).

Roberts Robert C. and W. Jay Wood. Intellectual Virtues: An Essay in Regulative Epistemology. (New York: Oxford up2007).

Rowland Richard . ‘Moral Error Theory and the Argument from Epistemic ReasonsJournal of Ethics and Social Philosophy 7 (2013) pp. 124.

Shafer-Landau Russ . Moral Realism: A Defence. (Oxford; New York: Clarendon2003).

Sosa Ernest . Knowledge in Perspective. (New York: Cambridge up1991).

Sosa Ernest . Judgment and Agency (New York: Oxford up2015).

Stanley Jason . Knowledge and Practical Interests. (Oxford: Oxford up2005).

Stapleford Scott . ‘Imperfect Epistemic Duties and the Justificational Fecundity of EvidenceSynthese 190 (2013) pp. 40654075.

Stapleford Scott . ‘Why There May be Epistemic DutiesDialogue 54 (2015) pp. 6389.

Streumer Bart . ‘Can We Believe the Error Theory?’ Journal of Philosophy 110 (2013): pp. 194212.

Wolterstorff Nicholas . 2010. “Entitlement to Believe and Practices of Inquiry” in Cuneo Terence ed. Practices of Belief: Volume 2 Selected Essays. Cambridge; New York: Cambridge up 2010) pp. 86117.

Williamson Timothy . ‘Knowledge as EvidenceMind 106 (1997) pp. 71742.

Zagzebski Linda . Virtues of the Mind: An Inquiry into the Nature of Virtues and the Ethical Foundations of Knowledge. (New York: Cambridge up1996).


Content Metrics

Content Metrics

All Time Past Year Past 30 Days
Abstract Views 93 93 12
Full Text Views 428 428 405
PDF Downloads 232 232 228
EPUB Downloads 3 3 3