Welcome to what should be a fun and insightful discussion of Spencer Case‘s “From Epistemic to Moral Realism” (generously made free access by Brill Online and the Journal of Moral Philosophy for this month). David Enoch has kindly contributed a critical précis, which appears immediately below. Please join the discussion!

Critical Précis to Spencer Case’s “From Epistemic to Moral Realism”

By David Enoch

I am delighted to start off the discussion of Spencer’s interesting paper, and I want to thank the PEA Soup people for inviting me to do this.

Spencer’s main claim – the one he is out to establish in this paper – is that if there are epistemic facts, there are moral facts (realistically understood). I accept this conditional’s consequent, so I’m fine with the conditional as well (I also accept its antecedent). But Spencer, of course, offers reasons to believe the conditional that do not depend on accepting moral realism – in fact, a part of the hope, I’m pretty sure, is to then employ the conditional as a part of an argument boosting the plausibility of moral realism. So in what follows I will present a quick overview of Spencer’s arguments, and then I’ll offer reasons to be suspicious of them. This criticism will be friendly, then, not just in the sense that we’re all engaged in a friendly cooperative effort to find the truth, but also in the sense that it comes from a fellow realist.

  1. Overview

Spencer rejects one argument for the conditional (if epistemic realism then moral realism), and then offers two arguments in its favor.

The argument Spencer rejects (pursued most influentially by Terence Cuneo in The Normative Web) is an argument by analogy. According to this argument, it can be shown that any reason to be suspicious of moral facts (realistically understood) is equally a reason to be suspicious of epistemic facts. Spencer is unimpressed for two reasons: First, he notes that this argument seems to presuppose that at least in metaethics, realism is the default position, so that arguments are needed to defeat it, not so much to establish it. Second, and now assuming an equally weighty reason to reject moral and epistemic facts, still it’s possible that there are weightier other considerations supporting epistemic facts than there are moral facts (like perhaps the observation that epistemic antirealism is potentially self-defeating in a way that moral antirealism is not). If so, the epistemic realist can still consistently reject moral realism – just like you can opt for the chocolate mousse rather than the ice cream, acknowledging that their equal price is just as much a reason to give up one as the other, but insisting that the fact that the chocolate mousse is nicer is a reason to prefer it (despite the price).

Both of the positive arguments advanced by Spencer make use of epistemic impurism, that is, roughly, the thought the pragmatic considerations, and indeed moral ones, encroach on epistemology, so that whether a belief is justified or amounts to knowledge may depend on what is pragmatically or morally at stake.

As Spencer does a great job clearly presenting his two positive arguments, I’m going to cite him here. First comes the 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.

The intuitive thought is simple enough – whether and how we should continue investigation (or settle on a belief, say) depends on how important the matter is, on what the costs are of continuing with the investigation, and so on. So moral considerations are in, and they affect epistemic status.

Then there’s the Evidence Possession Argument:

(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.

Here the notion doing work, as the name of the argument suggests, is that of the possession of evidence, or of available evidence. The evidence available to you is evidence that it’s in some sense sensible for you to put your hands on. But whether it’s sensible to get some evidence again depends on pragmatic and moral considerations. So again morality is in.

There is much more, of course, in Spencer’s rich paper (like discussions of evidentialism, mentalism, and more). But I think this will suffice to get us started, and also to render the critical points below intelligible.


  1. Evaluating Spencer’s rejection of the analogy argument

Perhaps it’s best to leave it to Terence to defend his Normative Web argument. Still, I’d like to make two quick points.

The first one is that I think Spencer is right that this argument presupposes that realism is the default view in metaethics. But I don’t think this is a problem, as realism clearly is the default view. No one would have even thought to look elsewhere if they hadn’t thought that realism (or some specific version thereof) faces serious challenges.

The second is about the dialectical point, which states, as you’ll recall, that even if the reasons against epistemic and moral realism are equally weighty, accepting the former and rejecting the latter may still be consistent, if the reasons for epistemic realism are weightier than those against moral realism. Now, Spencer is clearly right about some cases (as the price-of-desert example above shows). But I want to note that this line of thought does not apply universally. Sometimes, what we learn from “companions in guilt” arguments is something about the price, and then Spencer’s point clearly holds. At others, though, what we learn from a “companion in guilt” argument is that the initial objection was confused – sometimes, in other words, such arguments play the role of an undercutting (and not merely overriding) defeater. When they do, what we learn is that the relevant objections are just no good.

There is no general, content-independent way of determining whether an analogy functions in this undercutting way or not. So what we should now do, I think, is revisit Cuneo’s book, and see whether what he has to say about traditional objections to moral realism is that the epistemic analogy merely shows that they apply also to epistemic realism as a further price, or whether he thinks the analogy shows that they fail (and the answer may be different for different objections). If the latter, Cuneo is off the hook.


  1. Impurism, Terminological Issues that Don’t Matter, and What Does

As I already noted, both the Entanglement Argument and the Evidence Possession Argument rely on epistemic impurism. Clearly, then, if there are sufficient reasons in general to reject impurism, then these arguments are undermined. Now, I have my doubts about impurism, but I am not going to pursue them here. A part of the reason is that some of the relevant discussions (in the literature; not in Spencer’s paper!) seem to me annoyingly terminological.

I am a conservative on such matters. I like my epistemic-pragmatic distinction rather clear and tidy. So when the point is made that the pragmatic and moral costs of inquiry are relevant for whether or not the inquiry should be pursued – which is obviously correct – I like to think of this as them showing not that practical considerations are a part of epistemology, but that there are these practical considerations in the vicinity of epistemology. But what, except for terminology (and maybe a conservative temperament), is at stake here, really? Surely, the way the word “epistemological” is used is not something of substantive interest here.

OK then, if we are to avoid terminological issues, we can make progress, I think, in the following way: We can all agree that moral considerations are relevant to whether or not someone should engage in more inquiry. And we can all agree that there’s a sense of diligence relevant epistemically that is morally sensitive. And we can all agree that there’s a sense of “available” in which evidence that you can only acquire at unacceptable moral costs is not available to you. And if Spencer (and other impurists) wants to describe these phenomena in terms of the epistemic being encroached on by the pragmatic or moral, we should let them have the word “epistemic”. With all this agreed on, then, we should also agree that some epistemic facts are partly constituted by moral facts.

Isn’t this, though, all that Spencer set out to establish? I think not – for one thing, notice that there is something disturbingly trivial about this result. Sure, if we use our words in ways that render inquiry-relevant moral considerations epistemically relevant, then if there are epistemic facts of this kind, there are moral facts as well. No surprise there. So I think that Spencer wants more. To see this, think of what I will call “epistemic islands”.

Epistemic islands are parts of epistemology that are not pragmatically or morally encroached on at all. Assume (for now) that there are such parts to epistemology (more on this shortly). If so, while some epistemic facts entail moral ones, epistemic-island-facts do not. Those can be true even if moral realism is false. So it seems to me that what Spencer needs is the following disjunction: Either there are no epistemic islands, or, if there are, a position that is realist about them but not about the rest of epistemology is highly implausible. Or, to put the second disjunct differently: There’s no motivation for realism about non-island facts that is not equally a motivation for realism about island-facts. If this disjunction is false, there’s still the possibility of a view that’s antirealist about morality, as well as about the impure parts of epistemology, but that is realist about epistemic-island-facts. And if this disjunction is false, this kind of a view may be perfectly plausible, and Spencer may lose the crux of his analogy, and with it, ultimately, the reason to accept moral realism.

In a minute I’ll suggest some relevant epistemic islands. I think that they make the first line – rejecting the possibility of epistemic islands – highly implausible. Let me just note that perhaps some impurists – perhaps those attempting to do everything with a primitive concept of knowledge that is essentially tied to action, say – would like to deny the possibility of even just parts of epistemology that are pure. I find this kind of line implausible, but won’t say more about it here.

Let me conclude this section, then, with some suggestions for pure epistemic islands, such that (I want to suggest) being realist about them seems well-motivated even if realism about morality or the rest of epistemology has to be rejected:

(i) Even if you accept impurism about epistemic justification, or knowledge, you may still want to reject it (as I think most do) about credences. That is, perhaps whether a belief is justified depends on the practical stakes. But what credence we should have in p seems to depend on just the evidence. If so, credences are an epistemic island, and realism about their epistemic status need not commit you to moral realism.

(ii) Even if you accept impurism about diachronic evaluations – whether someone should gather more evidence, and so on – you may still be a purist about synchronic justification, namely, whether given a body of evidence a belief is justified or not. So this may be another island.

And there may be more islands of this kind. Spencer needs to do more in order to show that epistemic realism limited to islands is not an attractive position.


  1. The Evidence Law of Morality

A final point, one that is perhaps fueled as much by my current philosophical obsessions as by Spencer’s paper.

In discussing the Evidence Possession Argument, Spencer rightly notes that whether the smoking gun is available to the jury – in at least one sense of “available” – may depend on whether it can be retrieved in a morally permissible way. And he notes how if it cannot, they don’t have the relevant evidence available to them, which may affect what they are justified in believing.

Agreed. But I don’t think this is the relevant case for assessing the Evidence Possession Argument. For that, think of two cases in which the juries get to observe the guns, except that in one of them (but not the other) the gun was obtained immorally (and perhaps illegally as well). Now, it’s quite possible that the jury is not allowed to rely on the immorally obtained evidence in its decisions. But should the members of the two juries have different beliefs – and different credences? – depending on whether the evidence was obtained immorally?

It seems to me that the clear answer is no. And this seems even clearer if we leave the legal, institutional setting and ask whether what you believe of me should depend on whether or not the evidence that you can now consider was obtained immorally. This kind of consideration raises its own interesting questions. But this, of course, is a matter for another occasion.

24 Replies to “Spencer Case: “From Epistemic to Moral Realism”. Précis by David Enoch

  1. I wonder whether the Entanglement Argument identifies something that is peculiar to the epistemic. I would have thought that if there were moral facts, they would be pretty pervasive, or entangled in many different matters. For example, consider gustatory facts. One might think that a dinner is not a good dinner if making it has involved severe human suffering or severe violations of human rights. Someone who is inclined to accept nihilism about moral facts (as I am) should, I think, respond that insofar as some putative gustatory facts are entangled with moral facts, there are no such facts (i.e., there are no gustatory facts that are entangled with moral facts). But presumably, there are gustatory facts that are not entangled with moral facts, and such gustatory facts are unproblematic, from the moral nihilist’s perspective. I take this to be in line with one of the points that David makes in his précis. Likewise with respect to epistemic facts: There are no epistemic facts that are entangled with moral facts. But that does not mean that there are no epistemic facts. Some epistemic facts may not be irreducibly normative, in which case they are not targeted by one prominent argument for moral nihilism. For example, if the normativity of belief is functional, we might be able to give an account of reasons for belief according to which they are not irreducibly normative, as I argue in my contribution to the recent Metaepistemology volume (ed. McHugh, et al, OUP 2018).

    Regarding the Evidence Possession Argument, I agree that the jury does not possess the evidence that they can retrieve only by taking Smith’s family hostage. But I did not see why the fact that such an action would involve an “ethical cost” has explanatory force in this case. It seems to me that if the only way to retrieve a piece of evidence, E, is to take measures that are highly time- and energy consuming and uncertain in their outcomes (as seems to be the case with taking Smith’s family hostage), then one does not possess E. Whether these measures involve ethical costs seems explanatorily irrelevant.

  2. Thanks David for a nice introduction, and thanks Spencer for a nice paper!

    I have some questions about the “Robustness Argument”, which Spencer formulates as follows (p. 18):

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

    Spencer’s motivation for the first premise seems to be a claim along the lines of:

    (1*) If moral facts are less than robustly real, then these facts will be hostage to desire.

    The expression “hostage to desire” in turn seems to be cashed out in terms of counterfactual dependence, so that (1*) means something like:

    (1**) If moral facts are less then robustly real, then the moral facts would have been different if our relevant pro-attitudes had been different

    I have two thoughts about the dialectic here.

    First, error theory aside, many moral anti-realists at least take themselves to have views which do not entail that (1**) is true. Some that come to mind are: i) some versions of expressivism, where the idea is roughly that we evaluate the counterfactual situation using our actual attitudes; ii) “rigidified” versions of response-dependence theories, where, similarly, the moral status of actions in other possible worlds depends on the attitudes that we in fact have; and iii) sufficiently idealized versions of response-dependence, such as Michael Smith’s (where moral facts depend on attitudes that every possible agent would have in certain idealized situations) and perhaps Roderick Firth’s. (There are probably others too.)

    True, what Spencer says is just that “*most* forms of moral anti-realism would make epistemic facts dependent upon subjective mental states in unacceptable ways” (p. 19, my emphasis). So the idea might not be that (1**) holds across the board.

    But then my second thought is this: it seems that when it comes to views that are committed to (1**), this is already a devastating objection to them that does not have to do with their epistemological implications. It’s not plausible that torturing the innocent, say, would have been right if our attitudes towards it had been different (which is why many anti-realists have been so eager to reject (1**)).

    So then I wonder if the force of the robustness argument ultimately stems from the problem that has to do with (1**)? If that’s right, then it seems that anti-realists who can solve that problem — which they anyway have to solve — should be able to avoid the robustness argument too.

  3. It’s my pleasure to have an opportunity to participate in this forum. Here I will give my response to Professor David Enoch, whose comments are characteristically illuminating. (I saw Enoch’s critical precise in advance; I will reply to the other comments after I’ve had a bit of time to digest them). Here I will focus on Enoch’s central criticism, which I formalize as follows:

    Epistemic Island Argument
    1. If there are epistemic islands (i.e., epistemic facts not encroached upon by any moral or pragmatic considerations), then moral anti-realism can be motivated without any reliance upon encroached upon epistemic facts.
    2. There are epistemic islands.
    C1: Moral anti-realism can be motivated without any reliance upon encroached epistemic facts.
    3. If C1, then the argument from epistemic to moral realism is undermined.
    C2: The argument from epistemic to moral realism is undermined.

    I don’t object to 3, though it’s unclear how much my argument would be undermined if C1 turned out to be true. If all this means is that it some logically coherent version of moral anti-realism can be motivated without reliance upon entangled epistemic facts, then my argument could still have considerable force even if it were weakened to some degree. The version of moral anti-realism that could be motivated might not be the most plausible version of anti-realism. So my argument could make moral anti-realism less plausible without disproving it. As for 1, I think its plausibility depends on how big (and significant) the epistemic islands are.

    Although Enoch thinks that 2 is on firm ground, I think it’s probably the weakest link in the argument. The focus on apparently unentangled epistemic facts misses the larger point that the epistemic and moral domains are generally integrated. By way of illustration, politics and economics are intertwined. Certain events make this especially obvious, e.g., Herbert Hoover’s failure to be re-elected on account of the economy. The fact that a can of Coke-Cola in a particular vending machine costs 75 cents doesn’t seem relevant to politics. But it would be silly to think of this as an “economic island.” We know that politics and economics are thoroughly intertwined even though some economic facts, taken individually, seem not to have any bearing on political facts and vice versa.

    Similarly, the fact that some epistemic facts seem unrelated to moral facts doesn’t undermine the claim that epistemology and morality are generally entangled. The anti-realist might say: “My intuitions that some epistemic facts are completely unrelated to moral facts are stronger/clearer than the intuitions used to support encroachment.” But we wouldn’t necessarily expect to have an intuition supportive of moral encroachment for every epistemic fact. All epistemic and moral facts are part of the normative web, as Terence Cuneo puts it, but you can’t necessarily see that by zooming into a particular part of the web with a high-power microscope. We need to see those facts in the larger context.

    To return to the economic-political analogy: even the fact that a can of Coke-Cola costs 75 cents has some political relevance. If the American political system broke down completely, then it would have a much higher dollar price, or, if the dollar ceased to be a store of value altogether, none at all. The entanglement of some epistemic facts might be similarly easy to overlook.

    Seemingly isolated facts could be intertwined (though chains of inference) with other epistemic facts that are more obviously entangled with morality. Most, if not all, epistemic facts depend upon facts about the availability of evidence. I argue in the paper that this notion of availability is an impure epistemic notion. Hence all epistemic facts that depend on evidence availability would be encroached upon for at least that reason.
    What, then, of Enoch’s proposed examples of epistemic islands: credence and synchronic justification? Let’s consider each.

    Enoch writes: “perhaps whether a belief is justified depends on the practical stakes. But what credence we should have in p seems to depend on just the evidence” (emphasis mine). I want to ask: “what evidence?” Ostensibly, the evidence available to the agent at a time. Again, I direct readers to my argument that the relevant sense of “available” is impure. If that argument is sound, then credences are not epistemic islands.

    Second, I take both belief and credence to be psychological, not epistemic, notions. They are subject to epistemic assessments like justification, however, and moral considerations are relevant to these assessments. The intuition pumps I provide in the paper can be unproblematically restated in terms of credence rather than belief. For example, suppose that Smith the Creationist begins with a .85 credence that creationism is true. If he is not a diligent investigator and refrains from collecting easily available evidence that would lower his credence, then that is relevant to how his credence should be epistemically assessed.

    Enoch’s second example of a possible epistemic island is synchronic justification, sometimes called “propositional justification.” The first thing to say about this is that, for reasons I discuss in the paper, propositional justification is unlikely to be the only form of epistemic assessment. Even Earl Conee and Richard Feldman, who emphasize the significance of propositional justification, concede that there must be diachronic forms of epistemic assessment as well. If the anti-realist’s best response to my argument is to insist that all epistemic assessments must be synchronic, then my argument would have considerable force. Being forced to reject all diachronic epistemic assessments would be a significant bullet to have to bite.

    Second, I suspect that synchronic epistemic assessments are parasitic on diachronic epistemic assessments—in just the way that synchronic “evidence availability” is parasitic on diachronic notions on my account. Consider an analogy with “richness.” Mark Zuckerberg is “synchronically rich,” i.e., rich in a given instant. What that means, and why we would care to say it, is that Zuckerberg could, inter alia, go to the bank and make a large withdrawal. That could only happen over time, obviously. Likewise, I think what it means to be synchronically justified in believing a proposition is that you could, given your current mental state, blamelessly rely on it to perform certain actions, and arrive at new conclusions. If it didn’t mean this, then I don’t see why it’s an assessment we should care about.

    That doesn’t address all of Enoch’s criticisms, but I’m eager to hear from the other participants. Thanks again, Professor Enoch! Like both epistemic and moral facts, it’s been real.

  4. Circular Entanglement

    The reason the argument fails is that Case’s theory posits a troubling circularity. Here is my argument.

    (P1) If Case’s theory is true, then the epistemic norm (or norms) depend on a moral responsibility norm (or norms).

    (P2) If Case’s theory is true, then a moral responsibility norm depends on an epistemic norm.

    (P3) If (P1) and (P2), then there is a problematic circularity.

    (C1) Hence, if Case’s theory is true, then there is a problematic circularity.

    Premise (P1) can be seen in Case’s argument that doxastic states are epistemically better when they are accompanied by diligent inquiry or a lack of diligent inquiry that occurs because of a blame-lessening or –eliminating excuse such as ignorance, force, or impracticality. That is, epistemic norms (those that justify epistemic better-ness) depend in part on whether the agent is morally blameworthy for failing to further investigate something. In short, epistemic evaluation (whether in terms of justification or well-foundedness) depends on whether further investigation was required and whether further investigation was required depends whether someone would be blameworthy for failing to do so.

    Premise (P2) rests on the notion that blame rests in part on what a person knows. On one theory, a person is blameworthy for negligence only if he had a fair opportunity to know that what he did was wrong. On such an account, it is plausible that whether he is blameworthy for being negligent depends, at least in part, on whether there is an epistemic norm requiring him to recognize some fact. For example, a Rwandan genocidal leader who believes he is doing the right thing in ordering women to be terrorized and killed is morally blameworthy in part because he had access to the evidence that showed that what he is doing is wrong and is, thus, epistemically required to know that his order would be wrong. A similar thing is true for when someone is responsible despite being manipulated. Consider some hardline responses to Derk Pereboom’s four-case argument.

    On a second theory, a person cannot be primarily blameworthy for being negligent, that is, responsible on the basis of negligence alone. Rather, he can only be blameworthy if he knowingly performed an earlier act that made him responsible for a negligent act. On such a theory, a person is blameworthy only for morally akratic acts, that is, acts that he knew were wrong and did anyway. However, such a person is not responsible for all such akratic acts because he might know that an act is wrong, but be unable to avoid it due to an overpowering desire or manipulated decision. Also, he would not be morally blameworthy if he did not know how to prevent himself from acting akratically. On this last failure, a person is not morally blameworthy because he does not fail epistemically. He does not fail epistemically because his failure results from ignorance.

    In short, facts about moral blame depend on what a person knows or ought to know. If someone is morally blameworthy only for an akratic act, and thus what he knows, moral blame is tracking epistemic blame where a person is epistemically blameless for ignorance. If, instead, a person is morally blameworthy for what he does not know, but ought to know, moral blame depends on epistemic blame for failing to know something. What someone ought to know is at least in part an epistemic notion.

    Premise (P3) rests on the notion that moral norms and epistemic norms cannot rest on each other. To see, this consider which type of norm (or blame), epistemic or moral, is basic in terms of justifying the other.

    If epistemic blame is basic, then epistemic purism is true. Epistemic purism is the theory that epistemic assessments are determined entirely by factors that increase or decrease the probability of a belief being true from the standpoint of the agent. Examples include Bayesianism, reliabilism, and evidentialism. Such purism does not allow moral factors to sneak into epistemic norms.

    Case argues against purism. For example, he rejects evidentialism because it makes a person’s doxastic states correct when she has a justified true belief, but her belief does not depend on her justification. Consider, for example, Hillary Kornblith’s example of someone who has a justified true belief that q and justifiably believes p and if p then q, but does not hold q on this basis because she distrusts modus ponens. More generally, purism intuitively seems to capture propositional justification, but not doxastic or well-founded justification.

    If moral blame is basic, then epistemic norms may not be used to explain when a person is morally blameworthy. Intuitively, though, it seems that facts about the evidence available to the agent or, perhaps, the evidence that she was aware of, do explain, at least in part, when she is blameworthy for not having refrained from doing an act. Intuitively, the evidence-related norms are epistemic.

    A theory that one cannot be primarily morally blameworthy for negligence intuitively seems to depend on the notion that a person is not primarily epistemically blameworthy for negligence. This can be seen in that it is intuitively odd that one might accept that people are not morally blameworthy but are epistemically blameworthy for negligence or vice versa. The same is true for theories of moral blame that focus on, and only on, what the agent knew or was in some other way aware of.

    Here is the summary of the argument with the justification included.

    (P1) If Case’s theory is true, then the epistemic norm (or norms) depend on a moral responsibility norm (or norms).

    (P2) If Case’s theory is true, then a moral responsibility norm depends on an epistemic norm.

    (P3) If (P1) and (P2), then there is a problematic circularity.

    (C1) Hence, if Case’s theory is true, then there is a problematic circularity. [(P1)-(P3)]

    The following assumptions support premise (P1).

    Assumption #1a: Epistemic Failure. Whether a person is epistemically blameworthy depends on whether he did a diligent investigation.

    Assumption #1b: Moral Failure. Whether a person did a diligent investigation depends on whether he is morally blameworthy for failing to further investigate.

    The following assumptions support premise (P2).

    Assumption #2a: Moral Failure. Whether a person is morally blameworthy depends on whether he knows or should know something.

    Assumption #2b: Epistemic Failure. The relevance of what a person knows or the criterion for what a person should know depends on an epistemic norm.

    The following assumptions support premise (P3).

    Assumption #3a: Epistemic Norm Basic. If an epistemic norm is basic, then purism is true. Purism is false.

    Assumption #3b: Moral Norm Basic. If a moral norm is basic, then there is no explanation for the moral relevance of what an agent knows or should know.

    Case is likely to object that epistemic norms and moral norms depend on each other and that this is unproblematic. He might explain that this is similar to how consent depends on rights (valid consent waives a right) and rights depend on consent (a person has those rights he has not waived). He might also note that this is similar to how moral responsibility, particularly praise and blame, depend on what someone had a fair opportunity to avoid. What someone had a fair opportunity to avoid depends on when he would be blameworthy for not avoiding. Consider, for example, manipulation cases that happened in the distant past.

    Such an objection fails. The metaphysical justification here cannot be circular in this way or the metaphysical justification of moral blame would justify itself. Something cannot justify itself unless it is a foundational justification.

    This can be seen in that the above examples fail to illustrate what they are supposed to illustrate. Metaphysically, consent depends on rights. For example, consent is defined in terms of rights. Rights, though, do not depend on consent, even though they are affected by it. For example, rights are not defined in terms of consent. Also, if blameworthiness depends on what someone has a fair opportunity to avoid and what someone has a fair opportunity to avoid depends on what would make him blameworthy, then, in this context, either blameworthiness or fair opportunity does no work.

  5. Entanglement Misalignment

    There is an issue as to what an epistemic norm is. Let us assume that epistemic norms are understood in deontic terms that include epistemic duties and permissions. It is hard to understand what these might consist of other than unconditional moral duties or hypothetical imperatives. Similar to a legal obligation that is not a moral obligation, the idea of an epistemic obligation that is not a moral obligation is murky.

    If an epistemic norm just is a moral norm, then Case’s argument would need to focus more on why epistemic realism is true.

    If the epistemic norm is a hypothetical imperative, then Case’s argument does not succeed because unconditional moral duties likely will not entangle with epistemic norms in the way Case suggests.

  6. Thanks, Spencer.
    Not to belabor my points, but here are some responses:

    (i) I certainly agree that there may be connections that it’s hard to detect, and so that there may be apparent islands that are not islands after all. And I like the price-of-a-soda example. But there, of course, if pressed, I’ll be willing to step up to the plate and show the relations to the political system. So I really think it all depends on the plausibility of my examples for purported islands. If you can’t step up to the plate and show that they are not islands after all, then this line of defense is not available to you.

    (ii) Let me repeat the terminological point. In good spirit (!), I am here letting you use all words precisely as you want. So I have no problem when you say that the availability of evidence is partly moral. But this is in one sense of availability. And the question whether this is the only sense, or the sense relevant for my islands, is not settled by my willingness to give you the word “available”.

    (iii) Credences: Ok, then, here’s a case (which I take from a draft of a paper I’m writing with Levi Spectre, called “Statistical Resentment”): There’s an urn, with 30 red balls, and 70 black ones. One ball is randomly picked. What should your credence be that the ball is red?I submit (unsurprisingly): It should be .3. Furthermore, we have all the information we need for this. We don’t need to know about the stakes, or about a morally loaded notion of availability, or really anything at all. There is only one epistemically permissible credence here, and it’s .3, independently of anything else.
    I may be wrong about this, of course, but then you should step up to the plate – show me how morality encroaches on *this*.

    For now I’ll stop here (and I apologize if it will take me too much time to respond – I’m on Israeli time, and I hope to spend the night, well, sleeping.)

  7. Here is a brief response to Professor Jonas Olson.

    Thank you for your engagement. Your analogy with gustatory facts is interesting and not something that I had considered previously. This response is going to be off-the-cuff, then, so please take it in that spirit. I take you to be pressing the following argument:

    Olson’s reductio
    1.If you can infer moral facts from epistemic facts, then you can infer moral facts from gustatory facts.
    2.You can’t infer moral facts from gustatory facts.
    Therefore, you can’t infer moral facts from epistemic facts.

    Clearly, I’ve got to reject 1.

    Notice the ambiguity of “good dinner.” One sense of “good dinner” is clearly entangled with ethical facts. A dinner where the hosts treat the guests rudely isn’t going to be a good dinner in the sense that anyone wants to enjoy a “good dinner.” I think a dinner in which an unwitting enemy is served the cooked flesh of his children, as in some grisly ancient or Renaissance revenge play, most likely isn’t going to be a good dinner in this sense. The gustatory facts themselves don’t seem to be entangled, however. Though maybe if you were a vegetarian on ethical grounds you simply couldn’t enjoy veal no matter how well-prepared. Here, though, the entanglement is pretty clearly with psychological states, not ethical facts. If we understand “good dinner” so that ethical facts are presupposed, then it would be question begging. If we understand it the other way, then there would be no way to arrive at my conclusion.

    I think it would be bad news for me if there were a way of understanding epistemic assessments so that they were as obviously unrelated to ethical assessments, and another way in which they were, but only in a very question-begging way. My position is that there is no plausible way of understanding epistemic assessments so that they are all un-entangled. This is the right strategy for resisting my argument here: try to explain epistemic reasons in terms of mere probability or something like that. I think that if you try to do this, you’re going to have a very impoverished conception of epistemic reasons, one that seems to get a lot of apparent cases wrong. There’s obviously some reliance on intuition here. I don’t see how it can be helped: in order to investigate what epistemic assessments are, we have to have some intuitions about them. So it seems to me that epistemic and gustatory facts are relevantly different; hence I reject 1.

    As for the evidence possession argument, Olson writes (I’ll put it in quotation marks because I’m having trouble doing the indentations here):

    “It seems to me that if the only way to retrieve a piece of evidence, E, is to take measures that are highly time- and energy consuming and uncertain in their outcomes (as seems to be the case with taking Smith’s family hostage), then one does not possess E. Whether these measures involve ethical costs seems explanatorily irrelevant.”

    Note that even this apparently concedes pragmatic entanglement, which is no small matter. Presumably, it’s not the energy or the time themselves that matters to evidence possession, but the costs to the agent. I think once you concede that epistemic facts are entangled with pragmatic or practical facts, insisting on purity from moral entanglement seems arbitrary. Why, after all, should truths about my own well-being be relevant to facts about evidence position, but not truths about anyone else’s well-being? Also, the same sorts of thought experiments that suggest practical entanglement suggest moral entanglement. It’s counter-intuitive that an agent “has” evidence when bringing it to bear on her thoughts or actions would cost life or limb. It’s likewise counter-intuitive that she has that evidence when bringing it to bear would cost someone else life or limb (and she knows this). So I think once you concede that there’s epistemic-practical entanglement, you should concede that there’s epistemic-moral entanglement as well.

    That’s my very off-the-cuff response, but I hope we have an opportunity to discuss it in further detail.

  8. Hi Spencer,

    Very cool paper. I have some concerns that are similar to David’s (Hi, David!), but I might be closer to you than to him on the issue of interaction. My general scepticism about attempts to motivate, defend, etc. moral realism by appeal to epistemic realism is that I don’t quite see why anyone who is a realist about both would think that these two realisms require the existence of the same kinds of categorical reasons or standards. On its face, it seems that we might have multiple systems of norms that apply to all thinkers and all agents that nevertheless have reasons that, if they were to exist, would have significantly different properties. (I don’t want to get too hung up on the alleged differences because the point doesn’t require pinpointing the difference, really. It’s just a schematic worry. We know that people mean different things by ‘categorical’ and we might think that epistemic reasons and norms imply the existence of reasons that are categorical in sense one and we might think that moral reasons and norms imply the existence of reasons that are categorical in some different and possibly more robust sense. But if that’s how we understand these reasons and norms, it would be weird to think that the existence of the one would be evidence of the existence of the other.)

    But that’s why I think that the entanglement argument you offer is sort of cool. If there’s entanglement of the right kind, perhaps that’s some reason to think that the undermining strategy I’m envisaging (i.e., you cannot appeal to epistemic realism to motivate moral realism because epistemic realism requires only type-1 categoricity and moral realism requires type-2 categoricity). And like you, I sort of like entanglement views. Think about the Enkratic Requirement, the requirement to see to it that you don’t both believe that you ought to X and fail to intend to X (or fail to X). This seems like the kind of thing that many people accept and it’s hard to make sense of the idea that this is some kind of normative requirement if there’s not some kind of entanglement. (It seems to posit some direct link between the status of X-ing, which seems practical, and the status of believing something about X-ing, which seems epistemic.) Notice that in this example, though, we don’t need entanglement between the epistemic and the moral but the epistemic and some kind of practical requirement. And we know that there are practical requirements (e.g., instrumental requirements) that seem to apply to all rational agents even though they don’t have the interesting kind of categoricity that’s required by moral realism. (Or, so you might think.)

    And that’s a roundabout way of getting at what worries me about your argument. There is a view that I’ve seen kicked around in various quarters that says that the epistemic norms are the norms that govern things like belief. Then there are norms of inquiry or norms of reasoning that govern the way that we engage in theoretical reasoning. Now, maybe these norms have a kind of practical element to them because they have to do with the way that we pursue our epistemic ends and might also have to do with the epistemic pursuits we undertake in light of certain practical pressures that we’re under, but just because these norms of reasoning or inquiry aren’t (purely?) epistemic, it surely doesn’t follow that they have any moral import. And I’d think that norms of inquiry and norms of reasoning might cover the relevant cases (i.e., to explain our intuitive verdicts about availability of evidence and the way that inquiry connects to the epistemic status of a belief). One reason to think this is that it seems we can elicit the intuitions in cases that appear to have no moral dimension, we can explain the intuitions appealing to norms that look like norms of instrumental reasoning applied to the case of inquiring into a question, and we haven’t yet seen any reason to think that if we construct pairs of cases that are similar in all other respects but differ in terms of a connection to some salient moral consideration this changes much of anything about the epistemic side of things. (I’m assuming that the Cliffordian strategy of trying to show that the moral encroaches into every case of belief revision won’t work, but maybe this is the weak spot in the response I’m sketching.)

    [Apologies to posters from above. I might be repeating some points that have appeared before this post appears, but I suspect that some were approved and displayed in between my reading the post and the paper and my posting this comment. I think some of these points overlap with Jonas’s and Olle’s observations above. One thing I’d add to Olle’s concern is this. It seems that some of the paradigmatic requirements of rationality are both (a) desire independent in the sense that they’ll ban certain bad combinations of attitude whatever we happen to desire and (b) non-moral. So I’d think that the point about being hostage to desire won’t really favour the moral realist position being defended. Such requirements seem like they won’t be hostage to desire in a problematic sense, seem similar to epistemic norms, and seem to be non-moral.]

  9. Hi Olle Risberg,

    I don’t believe we’ve crossed paths before, but thanks for your thoughtful comments. Again, this is going to be somewhat off the cuff so bear with me.

    I want to reiterate what I was trying to do in that final section of the paper. At this point in the article, I took myself to have established that “if there are epistemic facts, then there are moral facts.” That doesn’t get me from epistemic to moral realism, though. In order to do that, I have to show that if there are realistically-construed epistemic facts, then there must be moral facts of the same level of robustness. If I didn’t argue for that last step, all I would have shown is that if epistemic realism is true, then moral nihilism must be false.

    I argue that if we accepted epistemic-moral entanglement and adopt some form of moral anti-realism other than nihilism, we would end up with some counter-intuitive implications for epistemic facts. Basically, they will be hostage to desire in ways that seem unacceptable. I think I am willing to provisionally sign on to your counter-factual characterization (though I might have to take that back, and retroactively decide that I really meant something else, if that gets me into trouble).

    The first part of your critique is that this isn’t necessarily true for all forms of anti-realism. You gave the examples of ideal observer theory and certain forms of expressivism. As for ideal observer theory, I recommend reading what Matthew Kramer has written about it in his book, Moral Realism as a Moral Doctrine. I think he makes it exceedingly clear how ideal observer theories like Smith’s do make morality hostage to desire in the way I describe. I can’t state it better than he does there.

    Let’s consider expressivism here, though. Some expressivists will deny that they are committed to the claim that if their attitudes were different and at variance with their current attitudes, then those attitudes would be morally laudable. They deny this because their expressions are indexed to their current attitudes in the actual world. This move has always seemed to me to be very unsatisfying, an illicit substitution of semantics for metaphysics. When we think of metaphysics, what is the ground for the expressivists current moral assertions? Is it not (at least in part) his moral attitudes? It seems to me that the expressivist has to acknowledge this. But saying both claims together seems very strange to me (and maybe contradictory, though I’m not sure):

    1.The ultimate metaphysical explanation for the wrongness of [insert whatever bad thing you like here] is my attitude toward it.

    2.If my attitudes toward [whatever it is] were different, its moral status would be exactly what it is.

    Another thought is this: couldn’t we use this kind of indexing to defend divine command theory? One of the objections to DCT is that if God willed [bad thing currently morally prohibited] then [bad thing] would be morally good, perhaps even obligatory. I don’t know if anyone has actually done this, but a sophisticated DCT defender might say: “No, those worlds in which God commanded [bad thing] really would be worse because [bad thing] is wrong.” This divine command theorist can simply adopt a semantic system that indexes the words ‘right’ and ‘wrong’ to the actual world. God’s commandments determine the meaning of those words, not just their extensions. So problem solved right? We no longer need worry that DCT is unacceptably arbitrary, right? I think probably not. If that move doesn’t work for DCT, I don’t think it works for the expressivist, either. But that’s just a sketch of an argument that I haven’t developed in detail anywhere. I don’t know if anyone else has pressed that objection to expressivism or not.

    Your last comment confused me because I’m not sure whether you intended to critique the conclusions of just this section or the entire project. It seems like it’s a critique of the whole project. Basically, who cares about biting epistemic bullets if we’re already committed to biting big moral bullets? In my recently accepted paper in the Australasian Journal of Philosophy, “The Normative Error Theorist Cannot Avoid Self-Defeat,” I address a similar question to this: why should the moral nihilist care about self-defeat when he’s already committed to saying so many counter-intuitive things? I give this more extended treatment there. For now I’ll say: there is something perverse about the idea that the more absurd your view appears at the outset, the more immune it is to any kind of refutation – or at least by reduction ad absurdum – because after all you’re already willing to say more ridiculous things than your opponent’s reductio would commit you to.

    I think a counter-intuitive implication either is grounds to reject a view or it isn’t. The fact that some interlocutors might be unflapped by those implications doesn’t matter to their epistemic significance. Sometimes I think the dialectical performance aspect of philosophy gets in the way of its ultimate goals. If you show that X has repugnant implication Z but that seems obvious, then the objection is going to be seen as stupid, question-begging and table-pounding. But if view Y has implication Z — and you can only show it by doing all sorts of fancy deductive backflips – then ladies and gentlemen we have a refutation! But how impressive the objection seems is irrelevant. Either the fact that a view has implication Z is a reason to reject that view or it isn’t.

    I think that if a metaethical position has the implication that epistemic facts are hostage to desire in the way that I describe, then that’s a reason to reject that view. Maybe we already have good reasons to reject those views, so that my conclusion is over-determined. I will not complain if anyone says this. Even if the conclusion is over-determined, I think that it’s interesting to know all of the reasons we have for accepting an interesting philosophical conclusion.

  10. Oh, Stephen Kershnar – I can’t resist addressing the jurispridential issue. Myself, I really don’t think there’s any plausibility at all – not even a tiny little bit – to thinking of legal norms as moral norms, or of legal duties as (perhaps legally created) moral duties, or of legal reasons as (perhaps legally created) moral reasons. I don’t think anyone would say differently, except if they have a jurisprudential axe to grind. But I digress…

  11. I have two follow-ups to what’s been said in comments so far. (I’m trying to use some formatting controls, but there isn’t any way to preview comments now — that was a very useful features that PEA Soup used to have — so I’m just hoping it works.)

    The first is a potential confusion that I really want to head off if possible. Spencer Case asked (and answered),

    When we think of metaphysics, what is the ground for the expressivists current moral assertions? Is it not (at least in part) his moral attitudes? It seems to me that the expressivist has to acknowledge this.

    That depends on what you mean by ‘the ground for … assertions.”
    If you mean

    (G) the metaphysical ground of the moral truths,

    then, no, that’s wrong. The epistemic basis for entitlement to make the assertions? What is expressed by the assertions? Then, yes, but… what follows?

    I think you must have meant (G), though, since that’s what you seem to be saying in

    1.The ultimate metaphysical explanation for the wrongness of [insert whatever bad thing you like here] is my attitude toward it.

    But no, there aren’t any versions of expressivism according to which (1) is true.

    I’d like to expound, also, on one part of David Enoch’s comments that I thought was particularly compelling, although I’ll develop it in a slightly different direction – I think David will largely agree with this development.

    Contextualists about knowledge attributions (and other epistemic lauds) have made a pretty good case that thresholds of justification are set by context, especially by practical considerations picked out by contexts. Let’s suppose, plausibly, that at least some of these are normative practical features. So, whether we can correctly predicate ‘knows’ of a person and proposition depends on these features. But here’s what does not follow: that whether the person knows the proposition depends on these practical normative features. What depends on those features is what precisely we are saying when we attribute knowledge.
    Compare what happens with expressions that are uncontroversially context-sensitive: quantifiers. Suppose I tell my colleague Josh,

    (MOST) Most students have taken logic.

    The conversational context sets the domain of ‘Most’. We’ve been talking about the philosophy concentrators at Brown, wondering what the demand for a course in contextualism next year would be. Whether what I said is true depends on what percentage of the philosophy concentrators at Brown have taken logic.
    Here’s what doesn’t follow: whether most students have taken logic depends on whether I am talking to Josh or to David Enoch. Rather, what precisely I said when I asserted (MOST) depends on that contextual feature.

    So the situation is this.

    Truth conditions of assertions using ‘most’: depends on who I’m talking to.
    Facts about whether most students have taken logic: don’t depend on who I’m talking to.


    Truth conditions of assertions using ‘knows that’: depends on practical normative features in context.
    Facts about whether people know things: don’t depend on those features.

    Now, this might be wrong – it may be that there are other plausible ways in which facts about whether people know things depend on practical normative facts. But my point is that it doesn’t follow from the very plausible sort of ‘practical entanglement’ that contextualists have argued for.

  12. Okay, so none of my html markup worked. Sorry about that. It still seems legible, just not pretty.

  13. David Enoch, you note that it is implausible that legal duties, norms, and reasons are in any way moral norms, duties, and reasons. I tend to agree with you, but worry about whether there is a value-free basis for the rules and rule-maker that, at least in part, constitute a legal system.

    My thought is that if there is a value-free basis for rules and rule-maker and if they justify the legal duties, norms, and reasons, then there are legal islands and an argument from the reality of these legal features to moral realism would fail. The same is true if legal duties, etc. are merely hypothetical imperatives.

    If there is value-laden basis for them, then the argument might seem to run from objective morality as necessary to explain these features of the law to there being objective morality. I then think the argument would depend on whether there is independent evidence that these features of the law really do exist. I don’t think there is such evidence, but am not confident about this view.

    I wonder whether epistemic realism is better off than the reality of legal duties, norms, and reasons. Perhaps the notion of credence and related examples are evidence of this.

  14. Hi Steve,

    I like your comments because they push me to address questions I’m still considering. I see I have my work cut out for me. But that’s another way of saying that I’ve got a robust research project. This reply addresses just the first part of your comments. I’ll get around to addressing others if I have time.

    I’m not convinced that mutual metaphysical dependence is itself a problematic thing to posit. It seems like there are all kinds of things that mutually depend on one another. Consider two vital organs within the human body: the heart and the brain. If one of them dies, the other dies.

    What is problematic, and what I think you’re alleging I’m committed to saying here, is that there are two things each of which is more fundamental than the other. If I were committed to saying that epistemic reasons are more fundamental than moral reasons, since moral reasons depend on epistemic reasons, and also that moral reasons are more fundamental than epistemic reasons that would indeed be bad: it would have the implication that each of these domains is more fundamental than itself! But that’s not the picture I imagine here.

    What are the exact parameters of the moral and epistemic domains? How, exactly, do they relate to one another? This is what I hope to investigate next. (I’m not keen on spending my entire career rebutting skeptics – at some point I’d like to make some positive discoveries about the nature of these norms). I tend to think that neither the epistemic nor the moral normative domain is fundamental. What is fundamental, on my view, is reason simpliciter, reason as such or “Just Plain Ought” (Owen McLeod’s term). The various normative domains that fall under this (e.g., moral, epistemic and practical) are just ways of classifying the many different kinds of normative reasons that apply to us. Where exactly the boundaries between these domains lie isn’t important. It might even be a matter of convention. What matters fundamentally is what reasons we have to act and think in various ways, and maybe what sorts of reactive attitudes we should have. How we classify the normative considerations we have isn’t so important. See my 2016 paper “Normative Pluralism Worthy of the Name is False” in the Journal of Ethics and Social Philosophy for a defense of this sense of ought. JESP is open-access and the article is easily searchable online.

    I don’t think that there are a plurality of different kinds of “blame” relative to different normative domains (e.g., epistemic blame, moral blame, practical blame). I just think that there is blame with regard to rationality-as-such. When we say that someone is morally blameworthy for some action, I think what we really are saying is something like: “This person’s action was bad all things considered in virtue of the fact that it ran afoul the moral reasons (i.e., reasons that we classify as “moral”) in a situation where he Just Plain Ought to have acted in accordance with them).”

    I’m not really sure I get the difference between the two forms of blameworthiness you describe. You write, “On one theory, a person is blameworthy for negligence only if he had a fair opportunity to know that what he did was wrong.” The other view is “In short, facts about moral blame depend on what a person knows or ought to know.” I think whether you ought to have known something and had a fair opportunity to know something seem pretty similar. Maybe I don’t know what all is being loaded into the word “fair” here. On the first view, if I could have searched for relevant evidence in the attic, but had no reason to suspect that I should do this then could that evidence affect later moral evaluations of my actions? Can you give a case where the two theories would diverge?

    What I would really like to find out is what is most basically assessable by rationality-as-such. Is it actions? Actions in combination with doxastic states? Purely mental actions? Mental actions of a certain type? I’m really not sure. Jonathan Spelman, in his dissertation, goes so far as to endorse the view that we ought to do whatever we believe we ought to do. That leads to all sorts of counter-intuitive conclusions, but I can see what leads him to them. There’s something repugnant about strict liability in basic morality: you’re morally on the hook for failing to do what in fact will maximize utility, for instance. But if you take that intuition seriously, it’s hard to see where the line should be drawn. Anyway, food for thought.

  15. Hi Spencer,

    Thanks for your thoughtful response. I’m a bit short on time here so apologies in advance for being a bit brief.

    Like Jamie said, most or all expressivists would reject the claim about the “ultimate explanation” of moral facts that you had in mind. The one who might accept that is the rigidified-response-dependence theorist. But she too would insist that it doesn’t entail the counterintuitive counterfactual claim (1**). (I haven’t been able to get access to Matthew Kramer’s book yet — I’ll try to do that.)

    One popular model here has been facts about color. It’s a familiar idea that color facts are in some sense ultimately explained by our responses to certain things, but it doesn’t look like the color of things would change if our responses to them would change. If somebody would endorse these two claims, perhaps along the lines of a rigidified view, I wouldn’t yet feel any strangeness corresponding to the strangeness you find in endorsing claims 1 and 2 together.

    What this leads me to think (though please correct me if I’m wrong) is that perhaps your thought is ultimately that moral and/or epistemic facts have a kind of objectivity (or: that they aren’t “hostage to desire”) that goes beyond the counterfactual characterization that I offered in my first post. I agree that this is an intuitive thought. But I’d just note two things. First, it remains to be seen how exactly this stronger form of objectivity should be understood. In fact, there are some recent arguments to the effect that not even robust realists can accommodate this idea (due to Justin Clarke-Doane, Shamik Dasgupta, and Matti Eklund, among others; I could give you more detailed references later if you want). Second, if I recall correctly, the argument in the paper only sought to show that the relevant facts should be objective in the counterfactual sense, not in this stronger sense. Maybe they should also be objective in the stronger sense. But I don’t think your arguments have showed this yet.

    Concerning my last comment, you’re right that it isn’t clear what exactly I had in mind. I agree with you that whether we’re justified in believing P can’t plausibly depend on whether supporters of P are willing to bite a lot of bullets. My question is rather along these lines: for those of us who already agree that error theory (as well as anti-realist views that entail (1**)) are problematic because of their first-order implications, does your argument provide a *distinct* reason to reject those views? I take it that even for a non-error-theorist, it’d be odd to reason like this: “error theory not only implies that my pain isn’t bad, but also that your pain isn’t bad — that’s two problems!” This doesn’t even look like a case of overdetermination. In some sense it’s just a single big problem. My question is whether the situation might ultimately be the same concerning the relevant anti-realist view, your epistemic objections, and the objection that focuses on (1**). But I realize that problem- and reason-individuation might probably be a very vague and context-sensitive matter, so I acknowledge that this question might not have a very precise answer. (Hence my struggles in formulating it!)

  16. There are other participants whose responses I am eager to get to, but Prof. Enoch is pressing a point that seems very important. Addressing it might be clarifying for others. So here’s a further reply to Enoch.

    David, You write:

    (iii) Credences: Ok, then, here’s a case (which I take from a draft of a paper I’m writing with Levi Spectre, called “Statistical Resentment”): There’s an urn, with 30 red balls, and 70 black ones. One ball is randomly picked. What should your credence be that the ball is red? I submit (unsurprisingly): It should be .3. Furthermore, we have all the information we need for this. We don’t need to know about the stakes, or about a morally loaded notion of availability, or really anything at all. There is only one epistemically permissible credence here, and it’s .3, independently of anything else. I may be wrong about this, of course, but then you should step up to the plate – show me how morality encroaches on *this*.

    I think I stepped up to the plate already in my initial response when I said both credences and beliefs are subject to epistemic assessments like justification. If justification is impure in the way that I argued, then both beliefs and credences are going to be entangled. Also, this example doesn’t seem to have anything to do with an agent’s credences per se since it could be stated in terms of beliefs, not credences, about probability.

    It occurs to me that maybe we are talking past each other here. I understand beliefs to be bivalent epistemic states and credences to be scalar epistemic states. If what you mean is that probabilistic relations are not entangled with morality, I will say, “sure, but probability itself isn’t a genuinely epistemic notion.”

    It seems to me that you are substituting a logic or math question for an epistemic one. In order for this to be an epistemic matter, someone’s doxastic states have to be involved. You stipulate that “There’s an urn, with 30 red balls, and 70 black ones. One ball is randomly picked.” When you say that we have all the information we need, I take it that “we” refers to the people participating in this discussion including you and me, *not the person in the scenario you describe.* How does that person know that there are exactly 30 red balls and exactly 70 black ones in the urn? Has someone told him this? Have they counted them? What reasons does he have to trust that the selection really is random?

    I don’t think that you can stipulate that this hypothetical person has all the relevant background knowledge without introducing points of entanglement. Your stipulations, given in abstraction of any practical stakes, keep numerous possible points of moral-epistemic entanglement offstage.

    Also, this is a synchronic example. This person might not totally trust his memory about what happened thirty years later. He might have to reflect on it a bit, check his journal or ask a witness he considers reliable in order to be sure about what actually happened. If what I’ve argued is on point, then how much verification he should do in order to know that his belief is justified will depend on what’s at stake.

    Finally, I reiterate what I said earlier: I don’t think that it’s critical for me to show that there are zero epistemic islands. The epistemic realist is presumably going to want to be a realist about many more epistemic assessments than your island examples. If that’s right, my argument is still forceful.

  17. Hi Clayton,

    I appreciate the comments. If I understand your concern, it’s that we can get most of the same mileage by positing rules of epistemic-practical entanglement that are apparently not rules of epistemic-moral entanglement, and that this undermines my case. One thought I have is that it’s not entirely clear to me how we are supposed to delineate moral and non-moral practical considerations. My view is that the delineation between these two domains isn’t particularly important and might be a matter of convention rather than a distinction that carves nature at the joints.

    Morality is thought to paradigmatically concern the well-being of other people, but I want to say that your own well-being is also morally significant even though self-interested considerations are often filed under “pragmatic” or “prudential.” So maybe some of what you have in mind as purely pragmatic encroachment I would consider ethical.

    The sort of reason I think it makes most sense to say are non-ethical are instrumental reasons. I’m actually skeptical of instrumental reasons. I think they are far less common in our practical thinking than many people seem to believe. Usually when I want something, there is some good that having it would confer upon me. So I think we have trouble delineating the reason I have because it would benefit me (or someone or something I care about) and the instrumental reason itself. When I think about cases of desires whose satisfaction confers no benefit upon the agent (or would even harm the agent), I don’t get any clear intuitive read that the desire itself creates much of a reason (though it might create a motive, which is something different). When I consider the reason I have to satisfy a desire I have to get something that’s good for me, I think my well-being is doing most, if not all, of the normative work.

    Your suggestion seems to be that the fundamental difference between these two domains has to do with the kind of categoricity involved. I think I would need to hear more about that to comment. Is your idea that non-moral “practical” concerns are entirely instrumental and hence desire-dependent? In any event, I attempt to motivate full-fledged moral entanglement by appealing to examples where moral things were at stake (e.g., obligations to others conflict with the agent’s desire to conduct an unfettered investigation). Now you could get similar results in cases where the agent’s own well-being is at stake, but that I think wouldn’t account for all of the intuitions I try to elicit.

    I think we can run a reductio ad absurdum against the view that epistemic assessments are pragmatically or prudentially entangled but not morally entangled. We could consider a pair of cases in which an investigation has to be called off for some self-interested reason and a case in which it is called off out of concern for someone else’s well-being. I think it would seem really strange to think that an agent’s belief can be justified on his current evidence if further investigation came at expense to himself, but not if it came at the expense of others. I guess that’s a logically consistent view, but it seems implausible to me.

  18. Hi Spencer, (and everyone else),

    As you know, I think that this is a fine paper. I will mention a few concerns I have (some of which you have heard from me many times already). Regarding the general strategy of arguing from epistemic to moral realism, I have never fully understood the appeal. That is because I have always found moral realism to be far more intuitively obvious than epistemic realism. This, of course, depends on understanding epistemic norms as a system of hypothetical imperatives not to be a form of epistemic realism. If at least some epistemic norms are of the form “if you want to get true beliefs/avoid false beliefs, then x” (where “x” might be “reason thusly”), then you might think that those norms are not realist in the relevant sense. I understand Cuneo to be making such a claim. It has always seemed to me that a lot of epistemic assessments can and should be reduced to statements about means-ends reasoning, where the end is some combination of acquiring true beliefs and avoiding false ones. I think that many, perhaps most, perhaps even all, such assessments are not morally entangled. Perhaps these are the epistemic islands that David is talking about.

    But what of other epistemic assessments, in particular the kinds of assessments (often regarding justification) that you persuasively argue are affected by moral judgments? The crucial point, as David says, is whether those judgments seem to be robustly realist. For example, consider our judgment that Smith the creationist is less justified in holding his belief when he leaves the lecture because he doesn’t want to hear contrary evidence than when he leaves the lecture because of a family emergency. Is that judgment robustly realist? And if so, does that then support moral realism? I would say that the judgment is indeed robustly realist, but that that fact doesn’t lend any (extra) support to moral realism, because the judgment itself depends on moral realism. That is, whether I am likely to give a robustly realist interpretation of that judgment about epistemic justification depends crucially on whether I am a moral realist. I suspect that those who reject robust moral realism (no need to explain here why minimalist attempts to claim the mantle of “realist” or “quasi-realist” are just an uninteresting shell game) would be quite happy to reject the robustly realist interpretations of the kind of epistemic justification assessments you argue to be morally entangled. Should that embarrass them? Sure, but to the same extent and for the same reasons that rejecting the robustly realist interpretation of purely moral claims like “factory farming is a moral horror” should embarrass them. Relatedly, consider two people who don’t believe that the meat they eat was subjecting to horrific suffering while alive. They were both read charming stories about happy farm animals when they were children. One of them has lived their whole life in an obssessively meat-centric culture, and has never even heard of vegetarians, let alone vegans. The other is a philosophy professor, who has many colleagues who teach and write about the immorality of eating meat. The philosopher has made (and communicated to their colleagues) a conscious decision to avoid learning about the treatment of animals on factory farms, because they are worried that what they would learn would cause them to be uncomfortable with their gastronomic choices (this is based, distressingly, on more than one actual philosopher I know). I’m comfortable saying that the relevant epistemic states of the non-philosopher and the philosopher are different in the levels of justification they possess. But I think this is clearly because the philosopher is morally at fault, for their conscious refusal to acquire information which is readily available (many of their colleagues would be more than happy to supply it). That is, the epistemic-sounding judgment that the philosopher is not justified in lacking the relevant true beliefs about the treatment of the animals they so love to eat just is a moral judgment.

    One way to put the line of thought that I’m exploring is that it divides the realm of epistemic assessment into the external and the internal. I have never understood why some epistemologists seem to want to argue for “internalism” over “externalism”, or vice versa. There are clearly different evaluations that we can and do make. Let a thousand flowers bloom. Or at least two. I’m not an epistemologist. But, I mean, how hard can it be? Matthias seems to do it quite well, as does Mark Heller, and Mylan Engel, and I’ve even been told that Clayton is a passingly good practitioner. Anyway, here’s a working hypothesis, that I’m sure will be shot down by someone who knows what they’re talking about (it’s the weekend, and I’m dabbling here): external epistemic assessments, which include David’s so-called “islands”), are all some variety of hypothetical imperative (or something interestingly related to such), and not morally entangled; internal epistemic assessments are often morally entangled, but are only robustly realist to the extent that the entangled moral judgments themselves are realist. The reason that the entangled judgments seem to be obviously non-trivially true is that the moral judgments seem to be obviously non-trivially true. If you can persuade yourself to accept some form of expressivism, emotivism, constructivism, or whatever about morality, you will be quite happy to give the same account of the status of the judgments about epistemic justification that seem to be morally entangled.

    One last point about evidence possession. I confess that I find it highly counterintuitive to say that the jury possesses the evidence (the knife) until they actually have it in the room. The mere fact that they could easily see it, if they asked for it, doesn’t seem (to me at least) to constitute their possessing it. I would say that they could easily, and morally unproblematically, possess it, whereas, in the case where they would have to take a hostage to see the knife, they could only with difficulty and morally questionable behavior, possess it. But the moral ease with which they could possess the evidence, doesn’t show that they do in fact possess it (before it is actually brought into the room).

  19. Dear Prof. Dreier,

    I appreciate your comments. I feel your pain when it comes to the formatting here. On the bright side, I’m pleased to see that we have at least this one point of agreement.

    You take issue with my saying that, according to expressivism, “the ultimate metaphysical explanation for the wrongness of [insert whatever bad thing you like here] is my attitude toward it.” You say, “But no, there aren’t any versions of expressivism according to which (1) is true.” I realize expressivism is your bailiwick, but I must demur. Olle Risberg points out that “most or all expressivists would reject the claim about the “ultimate explanation” of moral facts that you had in mind.” Yes, that’s true. But a theory can have certain implications even if all or most of the people who accept it deny that it has those implications.

    The kind of expressivism I had in mind when I wrote that is Simon Blackburn’s quasi-realism that allows him to say things like “Murder is wrong and it’s true that murder is wrong and it’s true that it would be wrong even if everyone on Earth thought otherwise.” This sounds like something a realist would say even though he understands moral statements to be expressions that do not have cognitive content.

    As I recall, on his view when I say “Murder is wrong” I am expressing my disapproval of murder. I can say “Murder is wrong is true” thanks to truth minimalism (i.e., ‘P is true’ just means ‘P’). And I can say it would be wrong even if no one thought so because there I’m expressing my disapproval of possible worlds in which no one has the attitudes that I actually approve of. Forgive me if I’m skimping on some details here, but I think that’s it in a nutshell.

    Now I don’t recall what Blackburn says about “moral facts” — whether he prefers to avoid moral fact talk or not. But I think as a naïve speaker of English that we should be able to say that “If X is true, then it’s a fact that X (or X is a fact).” Ok, so if we can say that the wrongness of murder is a moral fact, then what, on this view, would be the metaphysical explanation of that moral fact? Is it not the attitude that I am expressing when I say, “Murder is wrong”? It seems to me that it has to be. Blackburn attempts to address this in the appendix of Ruling Passions (OUP 1998), but I find what he says there to be unsatisfactory and evasive (e.g., “But there is no such meta-level question,” p. 311).

    If my attitudes are not the metaphysical explanation or basis for the moral fact I describe in this example, then what is? Does that fact have no metaphysical explanation or basis? Either nothing metaphysically explains this fact—which is mysterious—or something other than my attitudes do – which is inconsistent with expressivism—or my attitudes do. So I think we’ve got to say that my attitudes are the basis for this moral fact on that view. Alternatively we could say that according to expressivism there are no moral facts. In that case, the apparent sensitivity of epistemic facts to moral facts presents a challenge to the view, just as it does to moral error theory. Both are under pressure to drop epistemic facts, too.

    Now to your second point: You’re right that contextualism offers a competing explanation for the kinds of intuitions that I rely on to bolster my case. If it turns out that we can account for all of these intuitions by appealing to this semantic thesis, then positing epistemic-moral entanglement would be superfluous and my case would be undermined.

    I think if there’s any place where this essay could have been expanded, it would be to add a section dealing with the arguments for and against contexualism. I did cite Jason Stanley who argues for impurism in in his book Knowledge and Practical Interests (2005). I don’t add anything to existing critiques of the contextualism in my article except indirectly by making its competitor, epistemic impurism, seem more plausible. I think it would be an interesting result if it turned out that the moral anti-realist had to be committed to some form of epistemic contextualism. Since that is a controversial semantic view, my argument can impose a cost on the anti-realist if he doesn’t reject the argument at any other point.

    A small point: you talk about knowledge in your comments, but I’m primarily concerned with justification in the paper. I think someone who wanted to resist my conclusions by appealing to contextualism would have to be a contextualists about justification as well as knowledge. In fact, they would have to be contextualists about all epistemic assessments that appear to be entangled with moral considerations (e.g., well-foundedness). So the semantic commitments would be pretty substantial. Though some people seem to be willing to adopt any number of semantic commitments, provided that they can say everything the realist says.

  20. Hi Alastair,

    Yes, this debate is well-trodden territory for us, for sure. But it remains interesting.

    I agree that moral realism is more certain than epistemic realism. Imagine having to choose between two beliefs: 1) the beliefs of flat-Eathers are objectively unreasonable and 2) the Trans-Atlantic slave trade was objectively morally reprehensible. Now if someone produced an argument that showed that one of these two beliefs must be false, I wouldn’t hesitate to reject 1 and keep 2. This is not a tragic epistemic triage decision.

    (As an aside, I do worry that the Moorean thought behind it that can make any argument for moral realism impossible. There just aren’t that many things I’m more sure of than that some things are wrong, and that this isn’t just a matter of how people feel about them. Of course, that goes to David’s point earlier that it’s ok to assume realism is the default view. But I’ve always been dissatisfied with the defense only approach to defending moral realism (as is he, to judge by the arguments he’s produced for it)).

    Granting this point, I think this argument is worth defending for two reasons. First, we could still come to have greater confidence in moral realism by showing that epistemic realism entails it. Suppose that three extremely reliable witnesses testify that they saw something happen. Unless the event is something really strange or some other defeater is present, that gives you good evidence that it happened. Suppose that then a fourth witness who is moderately reliable corroborates the story. That should increase your confidence somewhat even though you already have stronger evidence for the conclusion. If the question is important enough, that extra confidence could be worth having. I think the truth of moral realism is pretty important.

    Second is a dialectical point. Presumably, moral anti-realists think that their anti-realism is rational, that their beliefs are epistemically justified. Now if I can show that the kind of rationality they’re committed to has broader implications than they realize, then I can hope to persuade them. You might think that the particular way I go about this isn’t promising because my interpretation of the cases that I give seem most plausible to me because of my acceptance of realism. That’s a good point and something that gives me pause.

    Note that I don’t immediately argue for a *robust* realist interpretation of those cases (I don’t get to robustness until the final section of the paper). I first just note that there appear to be relationships between epistemic and moral assessments without commenting on robustness. That might be something they haven’t recognized before. Could all of this easily be explained in terms of an instrumentalism that is compatible with moral anti-realism? Maybe, but a lot more needs to be said than that some ways of reasoning that are more truth-conducive than others. Philosophers have talked about all sorts of different epistemic assessments: justification, knowledge, well-foundedness, understanding and various intellectual virtues. These are notions that appear to be ethically encroached upon.

    To repeat something I said in my second reply to Enoch above, I don’t think pure probability statements are genuinely epistemic notions since they have no normative or evaluative aspect. That might push some moral anti-realists to say, “Fine, we can do without epistemic assessments.” But I think most will want to vindicate some of these thick epistemic terms. At that point, I can say, “Well, let’s take those concepts seriously, then. Let’s reflect on our intuitions about them a bit” and argue for moral encroachment. Later I argue that if they want to preserve their intuitions that epistemic intuitions are not hostage to desire, then they have to be moral realists. (Or they could remain anti-realists and own up to their epistemic anti-realism, which is also interesting).

    As to your last point about the evidence possession argument: do you also find it counterintuitive to think that some of my mental states count as evidence that I now have even though they are not immediately before my mind? You don’t want to follow Richard Feldman down the rabbit hole of denying this. He has correctly argued that there’s no principled difference between internal an external acts of investigation. Since he wants to avoid any practical entanglement whatsoever, he ends up with this absurdly narrow notion of evidence possession. Try to draw a different line between possessed evidence and non-possessed evidence that doesn’t commit you to at least practical entanglement.

    Thanks for continuing to think about this with me after all the time you’ve already invested!

  21. To all participants:

    Thank you very much for taking the time to read and discuss my work, and for engaging in such a thoughtful, civil manner. I have enjoyed this very much and would like to continue, but I’m afraid I have to disengage at this point. I’ve got an interview to prepare for and other obligations to attend to. Travis tells me that he’ll keep the comments open tomorrow so you’ll all have opportunities to respond to my comments. Just understand that I might not have an opportunity to reply. You’ve all given me much to think about. I hope we have the opportunity to continue the discussion in person at some point.



  22. The question, “what metaphysically explains the moral fact that F?” is not exactly transparent, so I think Allan Gibbard is right to distinguish different questions and answer them separately. In Thinking How to Live he says that moral facts are fully constituted by natural ones, but that there is another, more Fregean sense of ‘fact’ in which moral facts are distinct from natural(istic) ones and need no metaphysical ground.

    The mistake of thinking that expressivism makes the attitudes enter into the truth-makers, rather than being what’s expressed, has been made many, many times. One time played out in this very bowl in 2014:


    As that exchange shows, it is almost pointless (but only *almost*; the discussion there hints at an earlier iteration in which rational progress was made) to rehash the issue. So I think I’d better just refer to Mark Schroeder’s “Does expressivism have subjective consequences?” in Philosophical Perspectives 28 (2014).

  23. I know I’m late to the game, but I wanted to chime in too. I’m sympathetic to David Enoch’s point about “Epistemic Islands” undermining the argument, and I just want to expand on that.

    At least one influential version of “impurism” in epistemology–Fantl and McGrath’s, as developed in their book–is crucially committed to epistemic islands. Roughly, they think all the “thresholdy” notions in epistemology–justification, knowledge, reasonable belief–are functions of epistemic probability facts (which arepure), and practical considerations.

    More generally, I suspect that any kind of impurism on which we can systematically explain how stakes/morality/whatever “enroaches” on epistemic stuff will be structurally similar. That is, I suspect any such view will have to take the form of showing how purely epistemic facts interact with facts about stakes, options, etc., to give you “impure” epistemic facts. It’s telling that Stanley (on p. 87 of his 2005 book), whose official view is that epistemic probability is encroached, still offers a Fantl/McGrath style model–one where epistemic probability is not subject to encroachment, but interaacts with practical facts to generate knowledge facts–in illustrating his view. He takes it to be a convenient oversimplification, but my suspicion is that there’s no alternative to something structurally similar, unless you’re comfortable with a kind of mysterianism–one where epistemic and moral/practical facts are somehow entangled, but where there’s nothing much we can say about how that entanglement works. I talk about this a bit in section 4.2 of “Probability and Prodigality” (2013), for those who are interested.

    So ultimately, I think David Enoch is right that there are substantial, theoretically significant epistemic islands, and moreover that impurists have to agree, unless they want to be mysterian impurists.

Comments are closed.