Hope for the Evolutionary Debunker: How Evolutionary Debunking Arguments and Arguments from Moral Disagreement Can Join Forces

Abstract

Facts about moral disagreement and human evolution have both been said to exclude the possibility of moral knowledge, but the question of how these challenges interact has largely gone unaddressed. The paper aims to present and defend a novel version of the evolutionary “debunking” argument for moral skepticism that appeals to both types of considerations. This argument has several advantages compared to more familiar versions. The standard debunking strategy is to argue that evolutionary accounts of moral beliefs generate skeptical implications because they attribute those beliefs to factors that are unrelated to their truth. That strategy is vulnerable to “third-factor” responses, which invoke first-order moral claims to challenge the assumption that Darwinian factors and the moral truths are really unrelated in that way. In contrast, our version is immune to those responses, as it does not proceed via assumptions about how Darwinian factors relate to the moral facts. Instead, it focuses on what evolutionary accounts of moral beliefs have to say about the fact that people often reach divergent moral beliefs. The argument thereby illustrates how the debunking strategy can join forces with the argument from moral disagreement. The combination of those strategies presents, we think, a challenge that is more formidable than when they are considered separately.

1 Introduction

Evolutionary debunking arguments in ethics appeal to a Darwinian story about the origins of our moral beliefs to support skeptical conclusions about their epistemic status. The standard debunking strategy is to reach those conclusions via the assumption that the Darwinian factors that have shaped our moral beliefs are unrelated to their truth. This is supposed to show that it would be a sheer coincidence if those forces have led us to form true moral beliefs (which in turn is taken to support a skeptical view), even granted that there are moral truths.

By relying on the premise that the Darwinian factors that have caused our moral beliefs are unrelated to their truth, however, debunkers leave room for a powerful response. The reason is that whether those factors and the moral truths are unrelated is ultimately a first-order moral question. On some views about when an action is morally right or a person is morally good, they are indeed unrelated, but on others they are not. Hence, if a non-skeptic is granted not only the existence of moral facts but also the truth of some of her first-order moral beliefs, then she can invoke those beliefs to argue that there is a relation.

For example, consider our beliefs about moral rightness. On a simple evolutionary account, we tend to judge an action to be right in so far as it promotes the survival of ourselves or our kin. If one makes the moral assumption that this natural property also tends to make actions right, we can expect at least some level of alignment between our rightness-beliefs and the rightness-facts. So, given the moral assumption in question, a non-skeptic can deny that it would be a sheer coincidence if our moral beliefs were to correspond to the facts.

The idea that a non-skeptic can thus invoke first-order moral claims in response to evolutionary debunking challenge underlies so-called “third-factor” accounts of the reliability of moral beliefs (see, e.g., Brosnan 2011, Enoch 2010, Schafer 2010, and Skarsaune 2011; the example about survival is inspired by Enoch’s account). What is characteristic of those accounts is that they appeal to an external, “third” factor (such as the fact that an action promotes survival) which is taken both to (causally) help explain the belief that a given action is right, and to (metaphysically) explain that the action is in fact right.

Because of the availability of third-factor responses, the debate about debunking arguments has come to focus on the legitimacy of responding to skeptical challenges by relying on beliefs whose epistemic status is at issue (see, e.g., Korman and Locke 2020, Sinclair 2018 and Vavova 2014). And although things can be said on each side, debunkers do not seem to have the upper hand. The reason is that by denying their opponents the right to invoke the truth of their moral beliefs, the debunkers run the risk of committing themselves to forms of skepticism that go far beyond morality. After all, it is hard to see how one can vindicate the reliability of our beliefs in any domain without invoking the truth of some of them. Such “over-generalizing” implications are commonly agreed to render the debunking challenge less compelling.

Does this diagnosis make the debunking project hopeless? To argue that it does not is the aim of this paper. We shall pursue that aim by developing a version of the evolutionary debunking challenge that does not rely on the idea that the factors to which an evolutionary account attributes our moral beliefs are unrelated to their truth. By thus taking a different path, our version survives granting a non-skeptic the truth of her first-order moral beliefs and is out of reach of third-factor responses. Hence, it also avoids the over-generalization worries that arise if one insists that third-factor accounts are illegitimately circular. We take those features to be significant advantages.

The plan of the paper is as follows. In Sects. 2–4, we state and elaborate a principle about knowledge that is central to our version of the challenge. In Sect. 5, we go on to argue that an evolutionary account of morality supports thinking that moral beliefs violate this principle. Part of our argument relies on the fact that an evolutionary account does not only explain why we have the moral beliefs that we have but also why people often end up with radically different moral beliefs. Our version of the evolutionary debunking challenge therefore allows it to join forces with another prominent argument for moral skepticism, namely the argument from moral disagreement. In Sect. 6, we discuss in more detail the advantages that our version thus has over the standard debunking strategy, before making some concluding remarks in Sect. 7.

Before proceeding, three points about the dialectical geography are in order. The first point is that we will pursue our discussion under realist assumptions about moral beliefs and the facts they represent. Some debunkers think that one can avoid moral skepticism by adopting a different metaethical view, such as constructivism, but this contested issue will not be addressed here.¹

Second, although debunking arguments rely on complex empirical claims about the origins of moral beliefs, those claims are rarely subjected to detailed examination. Instead, the discussion focuses on what follows from the empirical claims that are put forward. Our contribution is no exception. What we seek to defend is the conditional claim that if certain empirical assumptions commonly made by moral skeptics about evolution and disagreement are correct, then moral skepticism looms. Since that conditional claim is the target of extensive criticism, including that from third-factor theorists, it would nevertheless be an important result.²

Third, there are different skeptical conclusions a debunker may pursue. One is that the evolutionary considerations undermine the justification of our moral beliefs. The conclusion of our version of the challenge, by contrast, is that there is no moral knowledge (even granted the truth of some of our moral beliefs). That said, we think this conclusion may have implications for the justification of moral beliefs as well, and will address this issue in Sect. 6.3.

2 Modal Accounts of Knowledge

Any attempt to derive skeptical conclusions from Darwinian considerations has to rely on an epistemic principle about what it takes for beliefs to constitute knowledge (or to be justified, etc.). The idea that underlies standard debunking arguments is that a belief does not amount to knowledge if it would take a sheer coincidence for it to be true. While our version of the challenge relies on the same general idea, we will focus on an aspect of it that has received less attention than alternative ones.

The idea that knowledge requires “non-coincidentally” true beliefs is familiar from the debate about Gettier cases, where it is held to explain why subjects in such cases lack knowledge. However, although many are attracted to that view, there is controversy about how it is best spelled out.³ One reason is that not all coincidences seem to have skeptical implications. If we come to believe some meteorological truth by accidentally meeting a meteorologist, say, then it is in some sense a coincidence that we have arrived at the truth, but this hardly excludes that we have knowledge. So how, more specifically, should the relevant concept of a coincidence be understood?

A popular approach is to do so in modal terms. On one attempt, knowledge requires that one’s belief is safe from error (see e.g. Pritchard 2005, Sosa 1999a, Sosa 1999b, and Williamson 2000). The basic idea is that a belief that p is safe just in case one could not easily have been mistaken with regard to p. Here is a characteristic formulation:

The Safety Requirement

S knows that p only if there are no nearby worlds in which (i) she does not have a correct belief as to whether p and (ii) she uses the same method of belief-formation as the one we actually used to determine whether p is true.

(ii) expresses a type of relativization that is typical for modal accounts of knowledge. It is motivated by the fact that what goes on in some nearby worlds is intuitively irrelevant to the evaluation of the subject’s actual beliefs. For example, suppose that a subject has obtained the true belief that there is a chair in front of her through using the “method” that consists in looking around. This belief plausibly constitutes knowledge even if there are nearby worlds in which the subject is blind (perhaps due to a freak accident that she has avoided in the actual world) and thus fails to discover the chair. By relativizing to methods, one can explain that world’s irrelevance. We shall return to some complications with this maneuver in Sect. 4. At this juncture, however, we shall focus on clause (i).

For present purposes, the crucial thing to note is that there are two ways in which a subject can fail to have a correct belief as to whether p (and thus also two ways in which a belief can violate the safety requirement). The first is that she believes that p even though p is false. The second is that she does not believe that p (perhaps because she believes not-p instead) even though p is true. Thus, if there is a nearby possible world where p is false but where she still believes p (and uses the same method), then her belief is not safe. But the same holds if there is a nearby world where p is true (and where she uses the same method) but where she does not believe that p. Earlier formulations of the safety requirement tended to focus only on the first way in which a subject could fail to be correct (see, e.g., see Sosa 1999a, Sosa 1999b, and Williamson 2000, 128). However, such formulations are now commonly taken to be too weak, as there are certain cases of coincidentally true beliefs that they fail to capture. For instance, if p is necessarily true, it is trivial that there are no nearby possible worlds in which p is false but the subject believes that p. But arguably, the subject’s belief that p can still be coincidentally true in the relevant sense—for instance, perhaps she has formed it simply through consulting a coin flip. Formulations of safety that focus on both ways in which one can fail to be correct have the intuitive result that the subject lacks knowledge in that case as well (assuming that there is a nearby world in which the subject fails to believe p, even though p is true, because the coin flip had a different result). That is why the stronger formulations have gained prominence (see, e.g., Clarke-Doane 2020, 108, who follows Pritchard 2009, 34).

3 The Adherence Requirement

When a belief that p does not violate safety in the second of the two ways just indicated (i.e., when the subject believes p in all nearby p-worlds where she uses the same method), it is said to be adherent. The idea that knowledge requires adherence is part of Robert Nozick’s influential “tracking account” of knowledge (1981). That account has been widely discarded due to problems with its other main element, according to which knowledge also requires “sensitivity” (roughly, that if p had been false then the subject would not have believed that p).⁴ The less-discussed adherence requirement has fared better, however, as is shown by the fact that it is entailed by recent formulations of the safety requirement.

The epistemic assumption which our construal of the debunking challenge invokes is a version of the adherence requirement. Interestingly, a similar idea can also be seen to underlie some of Darwin’s own speculations about his theory’s implications for morality:

If, for instance […] men were reared under precisely the same conditions as hive-bees, there can hardly be a doubt that our unmarried females would, like the worker-bees, think it a sacred duty to kill their brothers, and mothers would strive to kill their fertile daughters, and no one would think of interfering. Nevertheless, the bee, or any other social animal, would, in our supposed case, gain […] some feeling of right and wrong, or a conscience. (Darwin 1871, 73)

The idea here is not that we could have been morally mistaken because our beliefs could have been the same while the facts were different. Rather, the idea is that our moral beliefs could have been different from what they actually are (even if the moral facts remained the same) since evolution could have taken a different course. Later Darwinists have reasoned similarly. For instance, here is Michael Ruse and E. O. Wilson:

Suppose that, instead of evolving from savannah-dwelling primates, we had evolved in a very different way. If, like the termites, we needed to dwell in darkness, eat each other’s feces and cannibalise the dead, our epigenetic rules would be very different from what they are now. Our minds would be strongly prone to extol such acts as beautiful and moral. And we would find it morally disgusting to live in the open air, dispose of body waste and bury the dead. Termite ayatollahs would surely declare such things to be against the will of God. […] Ethics does not have the objective foundation our biology leads us to think it has. (Ruse and Wilson 1985, 52)

However, although the observations by Darwin, Ruse and Wilson are congenial with our version of the debunking challenge, they do not amount to full-fledged arguments for moral skepticism, even when they are combined with the idea that knowledge requires adherence. The reason has to do with the relativization to belief-forming methods. While it is hard to know how our cognition would have been structured if we had evolved like hive-bees or termites, it seems reasonable to suppose that the methods that we would then have used to form moral beliefs are not the same as those we actually use. Thus, due to clause (ii) in the safety requirement above, those worlds are irrelevant to whether our actual beliefs constitute knowledge. The same goes for the component of it that requires adherence.⁵

4 Methods and Positions

As we indicated earlier, however, clause (ii) raises some problems. Before presenting our version of the adherence requirement we shall elaborate and rephrase (ii) to avoid those problems. Reconsider the subject who knows that there is a chair in front of her even though there is a nearby world in which she is blind. The existence of that world does not seem to exclude that she actually has knowledge. Now, suppose that there is also a nearby world in which the subject is not blind but where her eyesight is considerably worse than it actually is. Arguably, that world is just as irrelevant as the first to whether she actually has knowledge. Nevertheless, if she in that world has tried (and failed) to determine whether there is a chair in the room by looking around, then there is no clear sense in which the method of belief-formation that she uses in that world is different from the one that she uses in the actual world. One can of course cook up a purely technical notion of a “method” to get that result.⁶ Nonetheless, we suspect that the relevant factors extend so far beyond those that are captured by the ordinary notion of a method that this could be misleading. For these reasons, we shall avoid using the notion of a method and instead invoke the one of a subject’s “epistemic position”.

We take a subject’s epistemic position, in relation to some claim p, to be defined by the features of her and her situation that affect her capability (either positively or negatively) of discerning whether p is true. There are at least three important categories of such features. First, a subject’s epistemic position includes abilities of different kinds that she has used to form the belief, such as her reasoning skills, her perceptual skills, and her “intuition skills” (if such skills exist). The second set of relevant aspects are features of her surroundings that influence her capacity to use those skills, such as the lighting conditions, the absence of disturbing noises, and the fact that she is or is not subject to various biases. The third category includes the background beliefs (if any) from which the agent has inferred the target belief.

Crucially, our notion of an epistemic position is thus relativized, in the sense that an item can be part of the subject’s epistemic position relative to one belief even if it is not a part of her position relative to another belief. Whether it is a part of her epistemic position relative to a given belief depends on its subject matter. If the belief is obtained solely through observation, for instance, the relevant aspects of her situation consist of her perceptual recognition capacities and the conditions which are relevant for their exercise. If the belief is inferential, one feature of the agent’s epistemic position is her reasoning skills, while another is the set of background beliefs from which the belief is inferred. The case of inferential beliefs raises further questions, which also arise for other modal theories of knowledge (see e.g. Roush 2012). However, we will set those complications aside in the present context and focus solely on non-inferential beliefs (for further discussion of adherence and inferential beliefs, see instead Risberg and Tersman 2019).

The definition of adherence that we shall adopt, then, is the following one:

Adherence: A subject S has an adherent belief that p if and only if she believes that p in all nearby worlds in which (i) p is true and (ii) her epistemic position, in relation to p, is at least as good as it is in the actual world.

However, the epistemic requirement that we would like to propose does not just state that a subject knows that p only if she has an adherent belief that p. The reason is that there are cases in which the adherence of the subject’s belief that p is ensured by factors that have nothing to do with her epistemic position in relation to p. For instance, suppose that the subject accepts some complicated mathematical theorem because of wishful thinking, and suppose further that her wishful thinking is so deeply rooted that it ensures that she has the same belief in all nearby worlds. Since wishful thinking is not a cognitive resource in a relevant sense, the modal robustness of her belief is, in this case, immaterial to her mathematical cognitive resources. It is therefore in the spirit of the adherence requirement to deny that the subject’s belief constitutes knowledge in this case (even if it happens to be true and she holds it in all nearby worlds in which it is true), and it would be desirable to formulate an epistemic principle that has this result. The most straightforward way to achieve this is to require not just that the belief is adherent, but also that its adherence is ensured by the relevant features of the subject’s epistemic position (rather than by wishful thinking or other similarly irrelevant factors). Hence, our version of the adherence requirement is as follows:

The Adherence Requirement: A subject knows that p only if she has a belief that p whose adherence is ensured by her epistemic position (in relation to p).

In what follows, we explain how evolutionary considerations suggest that our moral beliefs violate this requirement. As we stressed, we shall focus on non-inferential moral beliefs, as we take it that inferential moral beliefs constitute knowledge only if they are inferred (directly or indirectly) from some non-inferential moral beliefs that constitute knowledge.

5 The Adherence of Moral Beliefs: Evolution and Disagreement

The strategy we shall use is of the divide-and-conquer-variety. The point of departure is a distinction between two kinds of cognitive skills and abilities that could potentially ensure the adherence of a set of beliefs. Some of our skills are “domain-specific” in the sense that they can be used to acquire knowledge only, or at least primarily, in some specific area. An example of such a skill may be our ability to quickly learn languages at an early age. In addition, we have “general” skills that we use to acquire knowledge in many different areas. Hearing and deductive reasoning are two examples of such skills.

Now, if it can be shown that we have neither domain-specific skills nor general skills that secure the adherence of moral beliefs, it follows (given that the distinction is exhaustive) that those beliefs violate the adherence requirement and therefore do not constitute knowledge. And what we shall argue is that this can be shown. Evolutionary considerations suggest that we lack domain-specific skills which could ensure the adherence of our moral beliefs. However, they do not rule out that our general skills fail to make our moral beliefs adherent. In that case, we shall instead appeal to moral disagreement. This is how our reasoning allows the evolutionary debunking strategy and the argument from disagreement to join forces.

Our argument against the existence of sufficiently robust domain-specific skills in the moral domain focuses on the commonly accepted fact that Darwinian processes have only partially shaped the contents of our moral beliefs. The dispositions we have inherited define certain broad, vague limits within which our moral commitments may fall. What is excluded by those limits is, for instance, the view that it is an absolute moral duty to kill one’s offspring. Within the limits, however, other factors determine what moral views we, more specifically, embrace.⁷ One reason for that “leeway” is that the enhanced fitness of having moral beliefs is to some extent independent of their specific contents. The acceptance of moral rules in a community might enhance fitness by disposing its members not to free-ride or harm their offspring. However, many different moral rules could make them so disposed, and from the point of view of evolution (if one may say so), the details are not important.

Another reason why the Darwinian processes have promoted flexibility in our moral sensibilities is that the extent to which a set of moral beliefs enhance fitness depends on the circumstances in which the subject lives (where those circumstances may include the access to crucial resources). It is sometimes held that such differences explain, for example, why polyandry and polygamy are encouraged in some cultures but frowned upon in others. This context-sensitivity of the enhanced fitness that moral beliefs can bring also helps to explain why our inherited dispositions to form such beliefs allow for considerable leeway.

A third reason has to do with the value of the capability of adapting to new social circumstances. Groups change over time, through new members and new coalitions. To facilitate these changes, individuals need to be open to renegotiating their views about how to live. By stubbornly holding on to one’s initial ideas, without any ability to revise them, one might miss the opportunity of forming new alliances which may prove vital for survival.

If our remarks here are on the right track, they suggest that the Darwinian processes have worked against the emergence of domain-specific moral skills of the type that would provide adherence. For, while such skills would have steered us robustly towards certain specific moral beliefs—namely, the true ones—evolutionary processes have favored flexibility in our moral outlooks. Having such skills would, in many cases, simply not have benefited our ancestors.

We may now turn to our general skills. The reason why we need a separate argument in that case is that there has clearly not been a Darwinian pressure against the emergence of such skills. However, this is where moral disagreement comes into play. Our general skills clearly have some impact on our moral beliefs—for example, we use our reasoning skills to revise our moral beliefs upon finding that they are inconsistent. However, what the existence of certain types of moral disagreement illustrate is that those skills nevertheless do not secure the adherence of our moral beliefs.

It is commonly recognized that not all moral disagreements have skeptical implications. The fact that our moral beliefs are opposed by people who are less well equipped than ourselves in terms of reasoning skills, the lack of bias, access to relevant non-moral evidence, and so on, does little to show that our moral beliefs are not adherent. Such disagreements are as irrelevant as Darwin’s scenario in which we have evolved like hive-bees. The interesting cases are those of disagreement between people who are “peers”, in the sense that they are in equally good epistemic positions relative to the disputed issue. It is disagreements of that kind that skeptical arguments from moral disagreement have traditionally focused on, and the adherence requirement justifies this focus.⁸ For plausibly, if two people disagree about p even though their epistemic positions in relation to p are equally good, then neither party has an adherent belief about p, no matter whether p is true. Instead, what their disagreement reveals is that whoever is right about the issue at hand, their epistemic positions fail to ensure that each of them could not easily have had the belief of their opponent, even while the facts and their cognitive resources remained the same.⁹ More precisely, if p is true and the person who believes that p thus is correct, the fact that her opponent rejects p shows that their epistemic capabilities simply are not good enough to determine, in a sufficiently robust way, whether p is true. If p is instead false and the person who rejects p is correct, the same reasoning applies to her. Therefore, in a peer disagreement, neither party’s belief satisfies the adherence requirement (we develop this argument in more detail in Risberg and Tersman 2019).

The above reasoning shows that our general skills ensure the adherence of moral beliefs only if those beliefs are not contested by people whose general skills are at least as good as ours. And the point is that there seems to be a large number of cases where our disputants’ skills are not thus inferior. That is, it is plausible to suppose that our moral beliefs are very often contested by people whose general cognitive skills are at least as good as ours. While this is obviously an empirical claim, we shall not assess its merits but simply help ourselves to it in this context, since our aim is to show that there is room for a strong skeptical argument given the truth of the empirical conjectures that skeptics invoke (see Sect. 1).¹⁰ And, for the reasons noted above, this claim about the existing moral disagreements in turn excludes that our general cognitive skills ensure the adherence of the relevant beliefs.

Our version of the debunking challenge can thus be summarized as follows: Moral knowledge exists only if the adherence of some (non-inferential) moral beliefs is ensured either by domain-specific or by general cognitive skills. What the Darwinian considerations suggest is that we do not have domain-specific skills of the relevant type. The reason is that such skills would prevent the leeway and flexibility in our moral beliefs which are important preconditions for the survival value of having them. Moreover, the second possibility—that the adherence of our moral beliefs is ensured by our general cognitive skills—is ruled out by the existence of moral disagreement between persons whose general cognitive skills are equally good. Why do such disagreements not by themselves show that moral beliefs violate the adherence requirement? Because they leave open the possibility that the adherence of some of the conflicting beliefs is ensured by skills that are specific to the moral domain. What the skeptic needs is a positive reason to deny that there are such skills. And that positive reason is provided by the evolutionary perspective.¹¹ Thus, while neither the evolutionary considerations nor assumptions about moral disagreement by themselves show that our moral beliefs violate the adherence requirement (and thus fail to constitute knowledge), the combination of them do.¹²

6 Objections, Replies and Implications

In this section, we discuss how our argument can avoid some common objections to both debunking arguments and to arguments that appeal to moral disagreement. The objections we have in mind are those that appeal to third-factor accounts of the reliability of moral beliefs, the related over-generalization worries that arise if such responses are dismissed as illegitimately circular, and the response to arguments from disagreement that point to the fact that there are also cases of widespread moral agreement. We shall also discuss what the conclusion that there is no moral knowledge implies for the justification of moral beliefs.

6.1 Third-factor Accounts and Over-generalization Worries

As we noted in the introduction, third-factor responses to evolutionary debunking arguments invoke first-order moral beliefs to question the claim that the Darwinian factors that have shaped our moral beliefs are unrelated to their truth. Many debunkers criticize such responses by insisting that they beg the question (see, e.g., Fraser 2014, 471, Joyce 2016, 157–158, Street 2008, Sect. 6, and Street 2011, Sect. 6). On one version of this idea, the viability of third-factor presupposes that there are grounds for thinking that the pertinent moral beliefs are true, which is the very assumption that debunkers aim to refute.

The problem with this complaint is that it seems to rely on a very demanding epistemic principle, namely the idea that we are permitted to rely on our beliefs in an area only if we can explain their reliability without assuming the truth of some of them. The problem is that it is hard to see how our beliefs in any domain can escape skepticism if this idea is correct (for this point, see, e.g., Kelly 2008, Korman and Locke 2020 and Pryor 2004). For example, consider our perceptual beliefs. We do have an outline of an account of the reliability of such beliefs, which is based on our idea about how our sense organs work and have evolved. However, that account begs the question against perceptual skeptics in exactly the same way as that in which third-factor accounts begs the question against moral skeptics. For the only support we have for the assumptions that the account invokes comes from the very beliefs whose reliability it is meant to secure; namely, our perceptual beliefs. By ruling out third-factor responses on the ground that they are question-begging, debunkers accordingly risk committing themselves also to perceptual skepticism. And the same can be said, it seems, for most or all other domains of inquiry.

This faces proponents of the standard debunking challenge with a dilemma. They must either accept the legitimacy of third-factor responses, in which case their challenge fails, or dismiss those responses by adopting principles that risk leading to global skepticism. However, our version of the debunking challenge avoids this dilemma as it is out of reach of third-factor responses. For when confronted with our “argument from adherence”, it does not help the non-skeptic to invoke the (alleged) truth of her first-order moral beliefs, as the challenge this provides differs from standard versions. For example, even if it is true that euthanasia is impermissible or that polyandry is defensible, those truths do not undermine the argument against the existence of relevant domain-specific skills in the moral domain, since they do not change the fact that a significant amount of flexibility in one’s moral views is advantageous. Similarly, even if it is true that euthanasia is impermissible or that polyandry is defensible, those truths do not undermine our argument for thinking that our general cognitive skills also fail to ensure the adherence of moral beliefs. For since our first-order moral beliefs will often be disputed by subjects whose general cognitive skills are equally good—or so, at least, we have assumed—we may conclude that those skills do not ensure the adherence of either belief without taking a stand on the question of whether those beliefs are true. Hence, there is no need for an advocate of the argument from adherence to counter third factor-responses with principles that invite global skepticism.

The points just made show that our argument is not vulnerable to third-factor objections. However, they only amount to a partial response to the over-generalization concern, as they only undermine one possible argument to the effect that debunkers are committed to skeptical theses that go beyond ethics. We deem that there is some room for generalizability, but also that it is limited. In other words, it is likely that there are areas besides ethics where the relevant cognitive skills do not generate adherent beliefs but also areas where they do. Many perceptual beliefs are likely to be adherent, for example, since vision presumably counts as a cognitive skill (at least in favorable cases).¹³ By contrast, given a realist construal of aesthetic beliefs and the facts they concern, such beliefs may well fail to satisfy the adherence requirement. To determine whether the adherence requirement supports skepticism about a given domain requires a careful examination of the ways in which beliefs in that domain are formed, and we have not undertaken any such investigation.

It must be noted, however, that it is not always clear when the fact that an argument for moral skepticism can be extended to another area should be seen as a problem. Perhaps we simply should be skeptics about the other domain. Of course, if the area in question includes claims that are used as premises in the skeptical argument, then there are worries about self-defeat (cf., e.g., Cuneo 2007, Kyriacou 2016). Such a worry would arise if the argument from adherence were equally applicable to epistemological beliefs (including beliefs in the adherence requirement) as to moral ones. But, while this issue requires more attention than we can give it here, we think a case can be made for thinking that moral beliefs and epistemic beliefs differ with respect to their adherence due to possible differences in their evolutionary backgrounds. As we have stressed, the survival value associated with our disposition to form moral beliefs is to a great extent independent of which specific contents it ultimately results in. Many incompatible sets of beliefs can play the role responsible for the selection of the disposition. The same does not seem to hold for our disposition to make epistemic evaluations, whose main function is to steer us away from falsehoods. As Quine once put it: “Creatures inveterately wrong in their inductions have a pathetic but praiseworthy tendency to die before reproducing their kind” (1969, 126).

6.2 Moral Agreement

A common strategy among moral non-skeptics with regard to arguments from disagreement is to stress that although there is (peer) disagreement about many moral questions, there are also those that command near-universal agreement, such as, e.g., that pain is bad, that torturing babies is wrong, and so on. That observation has been held to diminish the skeptical worries the disagreement raises (see, e.g., Huemer 2016, Parfit 2011 and Smith 1994). For what the agreement might seem to suggest is that arguments from disagreement at best establishes restricted forms of skepticism, which target controversial moral beliefs but leaves others unscathed (for this position, see McGrath 2008).

However, whether an argument from moral disagreement is vulnerable to this objection depends on why the disagreement is supposed to generate skeptical implications. In our argument, the relevance of disagreement is due to the adherence requirement. Peer disagreement is simply a symptom of the more fundamental fact that the conflicting beliefs are not adherent. Since a belief can fail to satisfy the requirement even in the absence of this symptom, it is not immune to the challenge just because it is not disputed, not even if the subject also holds it in all nearby possible worlds in which its content is true. To satisfy the requirement, its adherence must also be ensured by the agent’s cognitive skills—otherwise, its status could be challenged on the same grounds as the belief in a mathematical theorem that is based on mere wishful thinking (see Sect. 4). This means that the existence of moral agreement spells trouble for the argument from adherence only if we have reason to think that the agreement obtains because our cognitive skills have ensured that it does. And the problem is that such explanations are less plausible than the alternative one which is offered by the evolutionary perspective our argument invokes. On the evolutionary account, there is agreement about the badness of pain and the wrongness of torturing children, not because we have cognitive skills that allow us to accept these views, but because rejecting them is beyond the broad limits on our moral commitments that our inherited dispositions set.

What makes the evolutionary account more plausible is that it helps to explain both why there is disagreement about certain moral issues and why there is a tendency towards consensus about others. With regard to questions concerning which there is no evolutionary leeway, we tend to reach agreement. When there is such leeway, however, we disagree because context-sensitive factors instead determine what different individuals end up believing. By contrast, the explanation of moral agreement which appeals to our cognitive skills leaves it mysterious why disagreement is so widespread about other moral questions. If we have skills that enable us to adherently believe that pain is bad, it is unclear why those capacities do not enable us to have adherent beliefs about the many other moral issues about which there is instead widespread disagreement.

Of course, these considerations do not conclusively refute an explanation of moral agreement which appeals to our cognitive skills. But the point is just that there is an alternative explanation of both moral agreement and disagreement that, we think, is superior. This is why our argument fares better also relative to the objection from moral agreement.

6.3 Moral Justification

We shall finally address the question of what implications our conclusions about moral knowledge might have for the justification of moral beliefs.¹⁴ The question is complicated by the dispute between internalists and externalists in epistemology. While the distinction between internalism and externalism can be drawn in different ways, our adherence condition on knowledge is at least manifestly externalist in the sense that it concerns modal and causal facts about a subject’s belief that are not always “accessible” to her. Internalists about justification typically doubt the relevance of such inaccessible factors. That said, as David Enoch notes, even though the debunking challenge (both his version and ours) “does start from an externalist perspective”, it can plausibly be “‘internalized’ […] even assuming that internalists are right about epistemic justification” (Enoch 2010, 423). For instance, it is consistent with internalism to hold that when a subject learns that her belief violates the adherence requirement, then that information undermines its justification. This accords with the following principle that Lutz (2020) takes to underlie Enoch’s construal of the debunking challenge:

Internalization: If F is a necessary, external condition on knowledge, and a subject’s belief that p lacks F, then, if the subject learns that her belief lacks F, the subject is no longer justified in believing that p. (Lutz 2020, 293, notation adjusted)

As Lutz notes, an attractive feature of this principle is that it explains why learning that one’s belief is coincidentally true if true at all tends to undermine its justification. For instance, if one believes that the time is 17:30 on the basis of looking at a clock, learning that the clock in question is broken is arguably undermining. And what one learns in this kind of case is exactly that one’s belief violates a necessary “external” condition for knowledge (i.e., a kind of anti-coincidence condition).

That said, the internalization principle also faces potential problems. For one, an internalist might suggest that in order for the discovery that a belief violates F to undermine the subject’s justification for believing that p, it is not enough that F is in fact necessary for knowledge—perhaps the subject must also believe (and/or have reason to believe) that F is necessary for knowledge.¹⁵ Moreover, there are some cases where the principle arguably fails to apply. It may well be reasonable to be highly confident that one’s lottery ticket will lose, for instance, even if one is also aware that one is not in a position to know whether the ticket will lose. On the other hand, this putative counter-example to the internalization principle is arguably due to features of statistical evidence and less-than-full belief that, we think, are not usually relevant for the justification of moral beliefs (for further discussion of knowledge and statistical evidence, see, e.g., Littlejohn 2020).

We will not attempt to sort out these complications here. We only wish to note that appealing to a principle along the lines just indicated is a promising way to argue from skepticism about moral knowledge to skepticism about moral justification.

7 Concluding Remarks

We have offered a version of the evolutionary debunking strategy which does not assume that the factors that have influenced moral beliefs are unrelated to their truth. We have furthermore argued that this is a crucial advantage as it puts our version out of reach of third-factor responses. Thus, we do not have to counter those responses by invoking epistemic principles that invite obvious over-generalization worries.

There is another feature of our version of the evolutionary debunking strategy which suggests that it is less likely to over-generalize; namely, that it pays more attention to the details of the explanations provided by the evolutionary approach. Some debunkers take the only relevant feature of that approach to be that it vindicates Gilbert Harman’s claim that the best explanations of our moral beliefs never cite their truth (Harman 1986). Accordingly, they tend to think that the evolutionary theory may be replaced in their challenge without loss by any other theory about the origins of moral beliefs that also vindicates Harman’s claim. For example, here is Joyce (see also Bedke 2014, Sect. 2, and Street 2006, 155):

The first thing to note about [the debunking argument] is that the evolutionary perspective is, strictly, dispensable. Were we to explain our moral beliefs by reference to, say, developmental and socialization processes, then, so long as these processes similarly nowhere imply or presuppose that our or anyone else’s moral judgments are true, the same epistemological conclusion could be drawn. (Joyce 2016, 125)

In our view, however, Harman’s claim provides a too slim basis for a skeptical challenge. A skeptic might appeal to it in defense against the objection that the mere fact that we have moral beliefs justifies thinking that they are true. But to turn that defensive argument into a more offensive one (for the conclusion that we lack moral knowledge or justified moral beliefs), the skeptic needs to make further problematic assumptions, such as that other types of justification somehow do not count. And then she once again risks committing herself to ideas that generalize too broadly by leading to skepticism also in other areas that deal with facts that seem to be causally inert, such as logic or mathematics.

In the argument from adherence, by contrast, the evolutionary perspective is not so easily replaced. That reason is that this argument does not just focus on Harman’s claim but on elements of the evolutionary approach that other theories about the origins of moral beliefs do not include, including those that pertain to moral disagreement. That allows us to argue more offensively for the view that there is no moral knowledge, via the idea that the evolutionary perspective suggests that moral beliefs violate the adherence requirement. Moreover, what the evolutionary perspective says about moral beliefs differs from what it says about the origins of beliefs in other domains. There is therefore no automatic way to extend the argument from adherence so that it entails skepticism about those domains as well. For example, whether we have cognitive skills that secure the adherence of our mathematical beliefs is a question our argument leaves open, even granted that the best explanations of those beliefs do not imply their truth.

Last but not least, when the debunking strategy is combined with the argument from disagreement, in the way that the argument from adherence does, we are able to avoid some of the most influential objections to both those strategies. The adherence requirement explains why the patches of near-universal moral agreement need not undermine the argument from disagreement, while the invocation of disagreement in the debunking challenge allows us to avoid third-factor responses to it. This interdependence is a crucial finding, as a fair assessment of moral skepticism—and, of course, of any other philosophical position—can only be done if one considers the totality of considerations that speak in its favor.