Moral knowledge and the existence of god

Abstract

In this essay, I argue that, all else being equal, theism is more probable than naturalism on the assumption that human beings are able to arrive at a body of moral knowledge that is largely accurate and complete. I put forth this thesis on grounds that, if naturalism is true, the explanation of the content of our moral intuitions terminates either in biological-evolutionary processes or in social conventions adopted for pragmatic reasons; that, if this is so, our moral intuitions were selected for their utility, not their truth; and that, if our moral intuitions were so selected, they are probably false. I defend the argument against three objections: first, that the argument amounts to a generic skeptical challenge; second, that ethical naturalism explains how our moral intuitions could have been selected for their truth; and third, that there is a pre-established harmony between the utility of moral beliefs and their truth-values.

Introduction

Consider the following rather obvious proposition:

Betraying a loved one for entirely selfish reasons is morally wrong.

This proposition represents a special kind of fact: a moral fact, a fact about what human beings (and perhaps other rational beings) ought not to do. You and I believe many propositions like this with near-unshakeable firmness. Indeed, we ordinarily think that we know these propositions to be true. I believe that our capacity for this kind of knowledge has enormous metaphysical implications; in particular, I believe that it gives us good reason to think that God exists. In this essay, I will argue that the probability that our moral beliefs are true is very low on naturalism but very high on theism, and that, in order to maintain confidence in our moral beliefs, we ought to affirm theism. My argument will not be deductive; I do not purport to offer an airtight demonstration of theism from self-evident truths. However, I do think that the argument is cogent in that most people, including naturalists, have very strong intuitive and empirical reasons to accept its premises.

Allow me to begin with several clarifications. First, I do not claim that theistic belief is a necessary condition of moral knowledge; it is perfectly possible for atheists to know moral truths. I claim only that the fact that humans have the capacity for moral knowledge lends support to theism. Second, no part of my argument depends on the thesis that moral facts are dependent on God in any way–it is meant to be compatible with a wide range of realist metaethical theories that naturalists might hold.

Third, in offering this argument, I presuppose what I will call metaethical realism. That is, I maintain that moral claims express truth-evaluable propositions and that many moral claims (probably the most important ones) have truth values that are not fixed by the attitudes of any individual or community toward them (Cuneo 2007: 20–51). If A claims that “Abortion is morally wrong under circumstances C” and B claims that “Abortion is morally permissible under circumstances C,” I take it that A and B genuinely disagree and that one of them is right and the other wrong. It just seems obvious to me that many moral claims–for instance, that inflicting harm for the fun of it is wrong, that discrimination on the basis of race or sex should not be tolerated, and that lying, cheating, or stealing for one’s own gain to the detriment of others is blameworthy–express true propositions, that anyone who denies them has mistaken moral beliefs, and that their beliefs would be mistaken regardless of what they or anyone else thought about the matter. I do not have the space to defend metaethical realism here; I simply note that the vast majority of us speak, behave, and reason as if morality has these features.¹ Anyone who does not share these assumptions will probably find my argument unpersuasive.

Fourth, I presuppose that moral intuitions are necessary to reliable moral belief-formation.² By moral intuitions I simply mean moral beliefs that are pre-theoretical and non-inferential. I do not hold that intuitions are the only epistemic grounds of moral beliefs, nor do I hold that they are self-evident (in the strong sense that understanding them is sufficient for believing them), incorrigible, infallible, or closed to revision. My argument is compatible with a reflective-equilibrium account of moral knowledge, as long as moral intuitions are a critical component in such an account. I think this presupposition is plausible, for moral theorizing without appeal to intuition is impossible; no viable, non-arbitrary ethical theory can be constructed without recourse to our pretheoretical beliefs about what is right and wrong, good and evil.³ Again, anyone who does not share this assumption will likely be unmoved by my argument.

The argument stated

Consider the following propositions,

(N) The universe was not created by an intelligence who had the will and the ability to endow humans with veridical moral intuitions,

and

(K) Some humans have arrived, or are able to arrive, at a body of moral belief that is largely accurate and complete.

By “largely accurate and complete,” I mean a body of moral belief that includes many more true beliefs than false ones and that includes beliefs (occurrent or not) about the vast majority of moral propositions that are relevant to the everyday moral lives of humans. Notice that N is implied by naturalism. Notice, further, that those of us who are not moral skeptics (and that would be nearly all of us) believe that K is true. I will argue that K is highly improbable on N and highly probable on ~ N, and therefore that anyone who accords high epistemic status to K has strong reasons to reject the former and affirm the latter.⁴ I take it that anyone with strong reasons to accept ~ N will have derivative reasons to affirm theism, since theism is the most philosophically defensible (or the only viable) metaphysic on the market that can accommodate ~ N.

Why should anyone hold K to be improbable on N? My answer to this question is inspired by “evolutionary debunking arguments” defended by philosophers such as Sharon Street (2006, 2008), Richard Joyce (2006, 2016), and Michael Ruse (1986, 2006).⁵ I take it that, if N is true, any plausible causal explanation of the content of our moral intuitions must appeal either to (i) biological evolution, (ii) the inculcation of moral conventions over time, or (iii) some combination of these two. I take it that these processes, possibly together with an assortment of other, minor factors, exhaust the plausible explanatory options open to the naturalist.⁶ If the naturalist can offer another plausible explanation of our moral intuitions which does not suffer from the problems outlined below, then my argument fails. But I do not think that there is such an account.⁷

If our moral intuitions are explained by any of these processes, then we have them only because they were useful to our progenitors. On the one hand, if Darwinian processes are responsible, then our moral-cognitive traits were selected only because they aided survival and reproduction. On the other hand, if our moral intuitions are the product of social convention, then those conventions must have been adopted either for pragmatic reasons or moral reasons (or for no reason at all, but that certainly won’t help). If adopted for pragmatic reasons, then they were selected only for their utility. If adopted for moral reasons, then the moral intuitions on which they were based must themselves be explained either by a Darwinian process or by convention. In order to avoid a regress, the explanation for these foundational moral intuitions must eventually terminate either in some biological-evolutionary process or in pragmatic reasoning, in which case those intuitions were selected only because they were useful.⁸ If N is true, then, we must ask ourselves: can we be confident that our moral beliefs are true if they were selected only for their utility? I do not think so; why should we expect moral beliefs that are useful to be true? Simon Blackburn (1993: 169–171) puts the point nicely:

Notice, too, the way the evolutionary success arises. Animals with standing dispositions to cooperate (say) do better in terms of other needs like freedom from fleas or ability to survive failed hunting expeditions by begging meals from others. No right, duty, or value plays any explanatory role in this history. It is not as if the creature with a standing disposition to help those who have helped it does well because that is a virtue. Its being a virtue is irrelevant to evolutionary biology… no natural story explains how the ethical sensibilities of human beings were made for the ethical properties of things...

Blackburn pinpoints the heart of the problem: if N is true, then moral truth had nothing to do with the formation of our moral intuitions, because it was entirely irrelevant to their usefulness and thus to their selection. Evolution did not condition us to care for our children, reward altruism, or punish treachery because it is right for us to do so, but only because doing so gave us a competitive advantage. For all we know, these actions are morally indifferent (or even morally wrong), and our beliefs about them are merely useful illusions foisted upon us by our evolutionary heritage. Indeed, if we had evolved differently within different familial and social structures, Darwinian processes and social pressures may have furnished us with entirely different moral intuitions. For instance, as Mark Linville (2009) points out, if we had evolved to be more like bees, females might have thought that they had a duty to kill males who were no longer fit for procreation, and if we had evolved to be more like wolves, we might have thought that one’s moral worth depends on one’s rank in a social hierarchy. This means it is highly unlikely that our moral intuitions correspond closely to moral truths. Such a correspondence would be a matter of sheer luck, since the processes that formed our moral intuitions were entirely disconnected from moral reality.⁹ As Baggett and Walls (2016: 202) put it,

At worst our moral belief-forming mechanisms are like darts thrown blindly, hardly likely to hit the small distant target of moral truth an ocean away, if there even is such a thing. This is the moral analogue of the clock that gives times produced by distant electrical signals entirely disconnected from the actual time.

As noted earlier, our moral intuitions play a foundational epistemic role in the development of our moral thought. Since, then, veridical moral intuitions are necessary for a veridical body of moral belief,¹⁰ it is very unlikely that K is true if N is true. So far, I have argued as follows:

• (1) If N is true,¹¹ then the truth values of our moral intuitions are irrelevant to their utility.

• (2) If the truth values of our moral intuitions are irrelevant to their utility, then they are irrelevant to their selection.

• (3) If the truth values of our moral intuitions are irrelevant to their selection, then it is highly improbable that our moral intuitions closely correspond to moral truths.

• (4) Therefore, it is highly improbable given N that our moral intuitions closely correspond to moral truths.

• (5) K is true only if our moral intuitions closely correspond to moral truths.

• (6) Therefore, K is highly improbable given N.

The phrase “closely correspond” in this argument is ambiguous. I invite readers to interpret it in whatever way they deem necessary for (5) to be true. I think that the argument will succeed on any reasonable interpretation.

My claim that K is highly improbable given N may appear to be too strong. One might argue that (3) should be replaced with the weaker premise,

(3’) If the truth values of our moral intuitions are irrelevant to their selection,

then the probability that our moral intuitions closely correspond to moral truths is inscrutable.

This would yield the more modest conclusion that the epistemic probability of K given N is inscrutable. I reject this amendment; (3) is true, and therefore K is, in fact, much more likely to be false than true given N.¹² To see this, consider the following case:

Myrtle’s Morality Machine: Myrtle, an eccentric inventor with a fondness for green Jell-O, has built a machine that can cause her husband, Bert, to spontaneously form, at a time, t, whatever complex of moral beliefs will most probably cause him at t to prepare green Jell-O for her. For example, Myrtle’s machine may cause Bert to believe at t that all husbands have a moral obligation to prepare green Jell-O for their wives whenever the clock strikes t. On days when Bert is upset enough with Myrtle that this belief would not be sufficiently motivating for him, the machine may produce in Bert the belief that wives deserve to be punished for aggravating their husbands and that serving Myrtle green Jell-O would be a punishment exactly proportional to whatever she has done to make him upset.

It is obvious that moral facts are irrelevant to the selection of Bert’s new moral beliefs in Myrtle’s Morality Machine, since they are irrelevant to the workings of Myrtle's machine. It is also obvious that the vast majority of moral beliefs produced by the machine (probably all of them) will be false, for presumably Bert rarely, or never, has moral reasons to prepare green Jell-O for Myrtle (let us suppose that Bert does not know that she is fond of green Jell-O and that she is too shy to tell him so), and so any belief of his that he has such reasons will surely be false.

But perhaps I have been unfair (after all, we all agree that there isn’t much morally interesting about green Jell-O). Consider the following, more liberal version of the example:

Myrtle’s Modified Morality Machine: Myrtle, caught up in a flurry of technological inspiration, modifies her machine such that for any state of affairs, x, which it is remotely feasible for Bert to bring about, her machine may be calibrated to produce in Bert the belief that he has an overriding moral duty to perform those actions which are most conducive to x. That is, the machine might motivate Bert to pursue any conceivable goal that is remotely feasible for him, from preparing green Jell-O to amassing a collection of 5”x8” purple notebooks to acquiring nuclear secrets and selling them to international terrorists. Every time Myrtle flips a switch, the machine randomly selects one value for x out of all possible values and produces x-conducive moral beliefs in Bert.

How probable is it that, after any given flip of the switch in Myrtle’s Modified Morality Machine, the moral beliefs subsequently generated in Bert correspond closely to moral truths? Obviously, this probability is not just moderately low, nor is it inscrutable; it is exceedingly low, so low that Bert, if he knew that his moral beliefs were so produced, would suppose most of them to be false. This gives us good reason to think that moral beliefs selected on some basis other than their truth are highly unlikely to correspond closely to moral truths; again, such a correspondence would be a matter of sheer luck. Since both biological evolution and the inculcation of social conventions are processes of this kind, if N is true, our moral intuitions are probably unreliable.

The naturalist might object that biological evolution and socialization are not analogous to Myrtle’s machine, since these processes are not random. This objector misunderstands the import of the thought experiments; Bert’s moral beliefs are never random (at least, not in the sense required for this objection to succeed). On the contrary, each time Myrtle flips the switch on her machine, the beliefs generated in Bert are calculated to achieve a predefined end (whatever x is). The problem with Bert’s beliefs is not that they are random but that they are selected based on whether they are x-conducive, rather than on whether they are true. Granted, it is possible that x-conducive actions might turn out, as a matter of coincidence, to be morally good, and so it is possible that a process which selects for x-conducive moral beliefs might produce moral beliefs that correspond to some degree to moral truths. But this occurrence will be exceedingly rare (for just consider the vastness of the domain of x and the sparsity of x-values which would produce this effect).

Note that, if it is highly epistemically improbable in Bert’s case that, after any given flip of the switch, the beliefs subsequently generated closely correspond to moral truths, then it is also highly epistemically unlikely that the naturalist’s moral beliefs closely correspond to moral truths. For human survival and reproduction, or whatever is the predefined end of biological evolution and convention-inculcation, is just one possible end among many and is no more epistemically likely than an x-value randomly selected by Myrtle’s machine to be in line with what is truly good. So (3) in the above argument seems to be true. Thus, the epistemic probability that moral intuitions produced by naturalistic processes are reliable is very low, not just moderately low or inscrutable.

I have so far ventured to show that the probability of K given N is very low. The probability of K given ~ N, on the other hand, is high. For if the universe was created by an intelligence of the kind described by ~ N, this intelligence would have good reasons to endow humans with veridical moral intuitions as well as the power to do so, whether by some miraculous act or by setting up the initial conditions of the universe in such a way that humans evolved to form reliable intuitions. (This note about the probability of K given ~ N may seem unnecessary, given what I have already shown about the probability of K given N. But I include it because, if K were improbable given N and given ~ N, then K would be improbable given a necessary truth, and we would be left with an argument for moral skepticism, not for theism.) The following premises complete my argument for theism:

• (G) K is highly probable on ~ N.

• (H) Therefore, all else being equal, ~ N is more probable than N given K.

• (I) If, all else being equal, ~N is more probable than N given K, then, all else being equal, theism is more probable than naturalism given K.

• (J) Therefore, all else being equal, theism is more probable than naturalism given K.

Once we have accepted (10), the real question is whether K is true–or, perhaps more accurately, whether we are justified in believing that K is true. That is an extraordinarily complex question, one which I will not attempt to answer in this essay. Suffice it to say that practically everyone believes K is true, since it is a necessary presupposition of moral reasoning, and that rejecting K is almost psychologically impossible. And if, in order to be rational, we must maintain probabilistic coherence among our beliefs, this means that practically everyone, in order to be rational, must give some credence to theism over naturalism on the basis of K.

Objections

There are many objections naturalists might raise against this argument, many of them drawn from the already expansive literature on evolutionary debunking arguments (see Wielenberg, 2016 for an excellent overview), though I think the debunking argument I offered above is immune to some of the more common of these. For reasons of space, I will limit myself to three pertinent objections. First, I will reply to the allegation that my argument amounts to nothing more than a generic skeptical challenge and therefore ought to be ignored. Second, I will reply to the objection that, if ethical naturalism is true, the truth values of our moral intuitions might be relevant to their selection. Third, I will reply to the claim, defended by David Enoch and Erik J. Wielenberg, that a plausible Darwinian account can be given of how moral intuitions might correspond to moral truths, even if their truth values are irrelevant to their selection.

Skeptical challenge or epistemic defeater?

First, the naturalist might object: “You are really just offering an argument for moral skepticism by placing an unreasonable burden on the naturalist to justify basic epistemological presuppositions. Thus, we have a right to ignore you, just as we have a right to ignore similar skeptical challenges to our capacity for empirical knowledge. All parties to moral discourse must presuppose the reliability of their moral intuitions; there is no special burden on naturalists to offer justification for this assumption.” This objection misunderstands my argument; I do not demand that the naturalist justify her belief that K is true. Rather, as a part of my case for theism, I have offered a rebutting defeater¹³ for this belief–that is, I have offered reasons, based on other beliefs that the naturalist holds, to think that, if she is right about naturalism, K is false, though it appears prima facie to be true. It is plausible that many moral beliefs, like many perceptual and memory beliefs, are prima facie justified even for those who cannot produce arguments for them. But even these basic beliefs are open to defeat if it can be shown that they were produced unreliably. This is not a general skeptical challenge, since it does not depend on a general demand that basic beliefs be justified. To illustrate the point: there is a difference between trusting one’s epistemic faculties despite inability to rule out the Cartesian demon hypothesis and trusting them despite belief in a Cartesian demon. To demand evidence against the Cartesian demon hypothesis is to offer a skeptical challenge. To argue that someone cannot consistently believe that his cognitive faculties are reliable while believing the Cartesian demon hypothesis is to offer a rebutting defeater. My argument is analogous to the latter challenge, not the former.

Ethical naturalism and the causal efficacy of moral facts

There is one strand of metaethical realism which might pose a challenge to a critical premise in my argument. Ethical naturalists hold, broadly, that moral facts are natural facts of an ordinary kind (Lutz & Lenman, 2018). Philosophers of this persuasion may hold that moral facts are entirely reducible to non-moral natural facts or that they are irreducible but supervene on non-moral natural facts. (My reply to this objection will be applicable to either construal.) David Copp (2008: 198–204), for instance, argues that moral facts of the form x is wrong are identical to “ordinary natural facts” of (roughly) the form x is prohibited by the moral code that best serves the needs of society. Obviously, if moral facts are natural facts, then moral truths are natural truths. And if moral truths are natural truths, they might turn out to be relevant to the utility, and thus the selection, of our moral intuitions. (On something like Copp’s account, this is especially plausible–the moral-natural truth that some action meets the needs of society certainly seems relevant to the selection of moral beliefs about that action.) So the ethical naturalist might contest premise (1). There has been lively debate in the philosophical literature over whether it is at all plausible to say, in concert with the ethical naturalist, that moral facts can feature in naturalistic causal explanations (see, for instance, Sturgeon, 1985, 1988; Harman and Thompson, 1996). For the sake of argument, I will concede that they can. I deny, however, that this is enough to rescue the ethical naturalist from my argument, for it can be shown that the truth values of our moral judgments are still irrelevant to their selection even if ethical naturalism is true.

If moral facts are natural facts which reduce to or supervene on non-moral natural facts, it is always an open question whether we are correct about which moral facts reduce to or supervene on which non-moral facts. For all we know, Copp is mistaken to think that the moral fact.

(F_m) Stealing is wrong is reducible to the natural fact.

(F_n) Stealing is prohibited by the moral code that best serves the needs of society.

Suppose that, unbeknownst to Copp, stealing is actually morally right, and although F_n obtains, F_m does not reduce to or supervene on F_n. Would the belief that stealing is wrong (let us call this belief “B_Fm”) be any less conducive to human survival and reproduction on this supposition than on the supposition that stealing really is wrong? No. B_Fm has precisely the same utility whether it is true or false, even on ethical naturalism. It is true that, if F_m does reduce to F_n, then when B_Fm is useful, it is usually also true. But it is not selected because it is true; it is selected for its usefulness, and it just so happens to be true. Notice that, if we suppose that F_m does not reduce to or supervene on F_n but rather reduces to or supervenes on.

(F_n’) Stealing is prescribed by the moral code that best serves the needs of society, then B_Fm is false, but it is no less likely to be selected. It is just as useful on this supposition as on the supposition that F_m reduces to or supervenes on F_n and that B_Fm is therefore true. So, even on ethical naturalism, the truth values of our moral intuitions are irrelevant to their utility, and thus to their selection. For all we know, moral facts do not supervene on the natural facts we think they supervene on, and our moral beliefs are wildly off the mark.

The ethical naturalist might object that reduction or supervenience relations are the same in every possible world, and so we cannot make counterfactual claims about what utility the belief that stealing is wrong would have if it were false, rather than true, because it couldn’t have been false (see Sturgeon, 1985). But this objection can be circumvented by stating the above argument in terms of epistemic possibility, rather than metaphysical possibility. If the utility of our moral intuitions is invariant across all epistemically possible distributions of supervenience relations between moral and natural facts, then the supervenience relations that actually obtain are irrelevant to the utility of those intuitions. And that means, given naturalism, that we have probably got the supervenience relations wrong.

Hold on, you might say: what gives us the right to supplant metaphysical possibility with epistemic possibility when reasoning counterfactually about causal relevance? I answer that good counterfactual reasoning–including counterfactual causal reasoning–often trades on epistemic possibilities that are not metaphysical possibilities. For example, while it is necessarily true that the morning star is identical to the evening star, it is possible to show that this is irrelevant to whether water boils at one hundred degrees Celsius by pointing out that water would boil at one hundred degrees Celsius whether or not the morning star and the evening star were identical. Similarly, all attempts to derive empirical predictions from identity claims in science depend on counterfactual judgments with impossible antecedents. For example, the counterfactual claim, “If water were not identical to H₂O, one mol of water vapor would not produce a mass reading identical to that of one mol of hydrogen gas and one half-mol of oxygen gas” has an impossible antecedent but was instrumental in the discovery of the chemical formula for water. Since my argument depends on a claim about the causal irrelevance of moral truth to natural selection, and since epistemic-counterfactual irrelevance is sufficient to demonstrate causal irrelevance, all that is necessary for my argument is that it is epistemically possible that we have the moral-natural supervenience relations wrong. And most naturalists are, I think, willing to grant that. (Indeed, it would be difficult not to without begging the question.)

Explaining correlation: enoch and wielenberg

The naturalist might seek to undermine premise (3) of my argument by offering an account of how a naturalistic process could result in a high correlation between moral beliefs and moral truths even if the truth values of moral beliefs are irrelevant to their selection. Both David Enoch and Erik J. Wielenberg have defended such accounts, to which I will now reply.

Enoch’s pre-established harmony account

David Enoch (2010) has developed a “pre-established harmony” account of moral knowledge (hereafter PHA) in response to evolutionary debunking arguments. He argues that, on the (plausible, in his mind) assumption that human survival and reproduction are moral goods, evolution will tend to push us toward evaluative judgments which promote good ends:

The fact that (roughly speaking) survival is a good pre-establishes the harmony between the normative truths and our normative beliefs… Survival (or whatever) is good; so behaving in ways that promote it is (pro tanto) good; but one efficient way of pushing us in the direction of acting in those ways is by pushing us to believe that it is good to act in those ways. And in fact, as we have just seen, it is good so to act. So the normative beliefs this mechanism pushes us to have will tend to be true. (2010: 431)

While Enoch admits that, on his view, the correlation between our moral beliefs and moral truths is probably not very strong, he claims it is strong enough to free us from skeptical worries.¹⁴

It seems to me that Enoch’s account fails, for three reasons: first, the initial premise of PHA, that human survival and reproduction are morally good, is indefensible for Enoch; second, even if this premise is true, it is epistemically improbable on PHA that the scope of moral values which we have evolved to recognize is adequately wide for our moral intuitions to be reliable; and third, if we accept PHA, we are forced into act utilitarianism of the most fiercely bullet-biting variety. I have tailored these criticisms to Enoch’s account, but I think that, with a bit of tweaking, they are equally applicable to other pre-established harmony accounts, such as those of Knut Olav Skarsaune (2011), Kevin Brosnan (2011), and Paul Bloomfield (2018).¹⁵

The normative assumption

The following normative assumption serves as the initial premise in PHA:

(NA) Human survival and reproductive success are at least somewhat good.

It seems to me that NA (and consequently PHA) is indefensible. For the realist cannot defend NA by appealing to his (or anyone else’s) pretheoretical beliefs about value, since my argument has called the reliability of such beliefs into question. But absent such support, there seems to be no good reason for the realist to give assent to NA rather than to the competing judgments that human survival and reproduction are morally bad or morally indifferent; NA is no more likely than either of these judgments to be true.¹⁶ Indeed, once her moral intuitions have been called into question, the naturalist might wonder whether any moral judgments at all are true, for as Joyce (2016) points out, an evolutionary account of moral cognition is fully compatible with error theory. Thus, the naturalist can have no confidence that there are any moral facts, let alone that NA is one of them. Indeed, if she concurs with the likes of Mackie (1977) and George Mavrodes (1986) that moral properties are “queer” in the sense that they are not at home in a naturalistic ontology, then the naturalist has positive reasons to deny that there are any such properties and thus to deny that NA is true.

Nevertheless, Enoch attempts to defend NA by the following argument:

And when it comes to creatures like us and our fairly close ancestors, the claim that their survival and reproductive success is of value gains much plausibility, I think, from the observation that survival (or some such) is at the very least good for the creature surviving, or for a close group of relatives, or something of this kind. Again, this may not be true of creatures in general (some creatures just do not have interests, and so presumably nothing is good for them). But when it comes to us and to creatures like us, this claim seems very hard to deny. Furthermore, it seems almost undeniable that there are close (if not obvious) connections between being good for someone and being good. (2010: 422-23)

The key move in the above argument is made in the final sentence: Enoch makes an inference from the goodness of something relative to some sentient creature’s interests to that something’s absolute goodness. But this inference is baseless without appeal to the pretheoretical judgment that the success of sentient creatures in promoting their own interests is mind-independently good.¹⁷ And, as noted above, such an appeal begs the question against my argument: it is irrelevant that our belief in the objective goodness of the flourishing of sentient creatures is “hard to deny,” since that belief cannot be presumed reliable in light of the challenge posed to the naturalist’s moral intuitions.

Folke Tersman (2017: 766) takes issue with this rejoinder in his own response to Enoch’s account: “…to require that a non-skeptic, when attempting to establish the reliability of our beliefs in an area, must not invoke any beliefs of the target kind is to impose a very strong condition; a condition that invites a more promiscuous type of skepticism.” We cannot expect the proponent of PHA to offer non-circular justification for his basic moral beliefs, Tersman argues, just as we cannot reasonably demand such justification from those who trust the deliverances of their senses. Tersman’s criticism is not applicable to my objection for the reasons given in Section 3.1. I am not offering a general skeptical challenge, but a defeater for the realist’s belief that NA is true, and one cannot overturn a defeater by appealing to one of those very beliefs which it defeats (see Vavova, 2015). I think this objection alone is decisive against Enoch’s account.

The scope of our value-recognition

Even if NA happens to be true, however, it seems to me that, if PHA is true, then the scope of moral values which we have evolved to recognize is too narrow for us to be at all confident in most of our moral intuitions. For on PHA, our moral beliefs have been shaped by a process that selects for beliefs conducive to only one moral good: the survival and reproduction of humans. But there may be many other moral values which we have not evolved to recognize and for which our stock of moral intuitions fails to account.¹⁸ Sharon Street (while making an unrelated point) notes:

The universe of logically possible evaluative judgements is huge, and we must think of all the possible evaluative judgements that we don’t see, from the judgement that infanticide is laudable, to the judgement that plants are more valuable than human beings, to the judgement that the fact that something is purple is a reason to scream at it. (2006: 133)

For all we know, any of these or even more intuitively absurd value judgments are true. This means that the scope of moral values that we have evolved to recognize might be exceedingly narrow, so narrow that there are many moral values of which we have no knowledge. This has (at least) two grave implications. First, for all we know, due to a failure to recognize them, we are neglecting moral values which we have a duty to promote. Second, for all we know, we are harming nobler moral goods in our endeavors to promote the lesser good of human survival and reproduction (for instance, by destroying plant life, which could very well turn out to be more valuable than human life, given that our moral intuitions are suspect).

If this is the case, our fund of moral beliefs is not very trustworthy, for we have knowledge of only a handful of moral values and no knowledge of what means to achieving those values are consistent with other, unrecognized goods. Indeed, many of our beliefs about such means are likely false. We are stumbling blindly through the moral universe; one wrong step could spell disaster. This is not the kind of security in normative judgment required for moral knowledge. Thus, PHA fails to rescue the naturalist from moral skepticism.

The inevitability of act utilitarianism

Now for my last objection: If PHA is true, Enoch will be forced to accept an extreme variety of act-utilitarianism. For, on his view, moral beliefs are warranted because they are the output of a cognitive system that reliably produces moral beliefs that are conducive to human survival and reproduction. This means that, in situations in which we know, on independent grounds, that a particular moral intuition is not conducive to human survival and reproduction, that intuition is defeated. (After all, we cannot justifiably hold beliefs that we know lack warrant.) What’s more, if, in such situations, we know on independent grounds that an intuitively immoral act would be conducive to survival and reproduction, the belief that the act in question is good has the same warrant that the intuition that it is wrong usually would–namely, that it is the output of a cognitive process that reliably produces beliefs conducive to human survival and reproduction. This leads to paradoxical consequences in cases like the following.

Imagine that you are being held hostage by a sadist who has threatened to take your life unless you torture a dozen innocent, very elderly persons for an extended period. In all likelihood, torturing these persons will result in the least harm to human survival and reproduction; the form of torture you would employ is unlikely to cause long-term damage, and most of your potential victims are unlikely to live much longer anyway. Furthermore, you are young, married, and far more likely than any of your potential victims to produce offspring before death, so preserving your own life would be conducive to human reproduction. I hope we all believe very strongly that, in this scenario, it would be morally wrong to comply with the sadist’s demands. But given the considerations mentioned in the previous paragraph, this belief of ours is unwarranted, for our moral cognition does not reliably produce value-conducive moral beliefs about scenarios of this kind. Obviously, we cannot act on moral beliefs which we know to be unwarranted; that would be irrational (and arguably immoral). In cases like these where the consequences of our actions are easily foreseeable, we ought to perform those actions which, upon deliberation, seem most probably conducive to human survival and reproduction, since deliberating about the consequences of our acts is a much more reliable way to form value-conducive moral beliefs in such scenarios. Thus, PHA collapses into act utilitarianism, at least in cases like this one.

Here is another way to understand the link between PHA and act utilitarianism: PHA depends on the assumption that “Survival (or whatever) is good; so behaving in ways that promote it is (pro tanto) good” (Enoch, 2010: 431). This amounts to the claim that the moral goodness of an action is derivative of that action’s utility in promoting human survival and reproduction. Therefore, PHA is true only if act utilitarianism is true. But obviously, act utilitarianism (at least of the kind implied by PHA) is false. Therefore, PHA is false. I conclude that Enoch’s account fails to give any significant plausibility to the claim that moral beliefs produced by natural processes correspond closely to moral truths.

Wielenberg’s epistemology of rights

Wielenberg (2014: 145) has attempted to show that our beliefs about moral rights are immune to evolutionary debunking arguments:

While there are various theories about the foundation of rights, it is widely agreed that if rights exist at all, their presence is guaranteed by the presence of certain cognitive faculties. The cognitive faculties in question are either the very ones required to form beliefs about rights or are closely linked to such faculties. Sufficiently cognitively developed creatures that are products of evolution will possess moral barriers (if such barriers are real) and will also be disposed to believe that they have such barriers. The very cognitive faculties that lead such beings to believe that they possess moral barriers also entail the presence of those very barriers… In this way, the relevant cognitive faculties are responsible for both moral rights and beliefs about those rights, and so the cognitive faculties explain the correlation between moral rights and beliefs about those rights.

(By “moral barriers,” Wielenberg means trans-culturally recognized moral rights against injury.) Wielenberg proceeds to invoke this epistemological story against debunking arguments presented by Street and Joyce.

Wielenberg’s account is open to at least two objections. First, his foundational assumptions about rights are indefensible for the same reason that Enoch’s normative assumption is indefensible. His story sounds initially plausible; at least to those of us from liberal western backgrounds, it seems obvious that any agent possessed of rationality advanced enough to entertain moral propositions has rights. But no appeal can be made to our intuitions about this matter, since the naturalist has a defeater for these intuitions. For all the naturalist knows, there are no such things as rights, or, if there are such things, they do not obtain when she supposes that they do, and her beliefs about rights are nothing more than useful illusions generated by biological evolution and social conditioning. Indeed, on naturalism it seems improbable that our theories of rights just so happen to line up with moral reality–if they do, we have made some pretty lucky guesses. So Wielenberg cannot appeal to the consensus among rights theorists without begging the question.

Second, Wielenberg’s account, even if true, only rescues a small subset of our moral beliefs. For we have many, many moral beliefs that are not reducible to beliefs about moral barriers. Consider, for instance, the belief that a life lived in love and self-sacrificial service is more valuable than one spent in isolation, hate, greed, and self-absorption. This moral judgment is not entailed by any judgments about moral barriers (or at least not by any plausible ones); someone might live the latter kind of life without ever so much as threatening anyone’s rights. So Wielenberg’s account fails to restore confidence in such moral judgments. This defect is lethal, for no account of moral knowledge is adequate if it fails to accommodate our beliefs about what lifestyles are morally better than others. Similarly, Wielenberg’s theory cannot restore confidence in our beliefs about special obligations–for instance, to friends or family–since the presence of cognitive faculties capable of producing beliefs about rights does not guarantee the presence of such special obligations (not every moral agent you meet is a friend or kinsman). Further, his account fails to rescue our beliefs about the moral badness of abusive or otherwise irresponsible behavior toward animals or other non-moral agents, since those agents do not have the cognitive ability to form beliefs about rights and therefore are not guaranteed to have them. I could continue to multiply examples, but the force of the objection is this: even if Wielenberg’s account happens to be accurate, our moral knowledge is so truncated as to offer no answers to many of our most pressing questions about what kind of people we should be and about what moral obligations we have. Thus, both Enoch’s and Wielenberg’s challenges to premise (3) fail.

Conclusion

I have argued that, if the universe was not created by an intelligent being with the will and power to endow us with moral knowledge, it is highly improbable that we can arrive at a body of moral belief that is largely accurate and complete, since on naturalism the truth values of our moral intuitions are entirely irrelevant to their selection. If we are ill-disposed toward a radical moral skepticism, then, we have powerful reasons to affirm theism or something like it. I do not, of course, claim this argument establishes full-blooded classical theism; additional arguments are necessary for that. I do, however, claim to have shown that naturalism is unable to account for our moral knowledge and should not be affirmed by anyone who trusts her moral intuitions. Since most of us would, I think, very much like to trust our moral intuitions, we ought to reject naturalism accordingly, and once we have jettisoned naturalism, theism seems like a very promising–in my view, the most promising–alternative.