1. The Darwinian Dilemma

In her (Reference Street2006), Sharon Street famously argues that it is much more likely that our moral views have been systematically shaped by evolutionary forces than that we have grasped moral realist-style moral truths. By moral realism, Street means the conjunction of the theses that (i) moral truths are mind-independent, (ii) our moral beliefs aim to represent them truly, and (iii) our beliefs tend to amount to knowledge. Her idea is that moral realists who hold (i)–(iii) cannot explain the moral knowledge they think we have.

Street starts out by arguing that our moral beliefs largely are shaped by evolutionary forces. If so, there is either a relation between the evolutionary influences on our moral attitudes and these mind-independent moral truths that is likely to make us track the truths, or there is not. This generates the dilemma.

With respect to the first horn, Street then argues that if there is no relation between the forces that affect our moral attitudes and the mind-independent moral truths, we are almost certainly off track from them. Evolutionary forces select for what is adaptive, not for moral truths. With respect to the other horn, Street adds that if there is a relation here, evolution must have selected for our ability to track moral truths, which is inconsistent with a good scientific explanation of our attitudes. This is because our best interpretation of the science supports an adaptive link account of our moral attitudes. On this view, our moral attitudes ‘forged adaptive links between our ancestors’ circumstances and their responses to those circumstances, getting them to act, feel, and believe in ways that turned out to be reproductively advantageous’ rather than track mind-independent moral truths (Street Reference Street2006: 127). Granted, what is challenging about the dilemma is disputed (cf. Bedke Reference Bedke2009, Reference Bedke and Shafer-Landau2014; Enoch Reference Enoch2010, Reference Enoch2011; Lutz Reference Lutz2018; Vavova Reference Vavova2015). Some suggest that it is a version of the Benacerraf-Field challenge, meaning that it seems to be a cosmic coincidence that we know anything about causally inert moral properties, for it is unclear how our attitudes link up with them. Others think evolutionary explanations of our moral beliefs serve as an undercutting defeater. I prefer the former interpretation, but the argument below can be reformulated to fit either.

On either horn, our moral attitudes are therefore unlikely to have much to do with mind-independent moral truths. If we then add the common realist premise that if we cannot know the moral truths, then mind-independent moral realism is false, it follows that we should give it up.

2. Some Terminology

There is, however, some controversy about which views are challenged by Street's dilemma (cf. Copp Reference Copp2008; Street Reference Street2008). To take aim, I shall therefore distinguish between non-naturalist moral realism and moderate metaethical views. Expanding on Street's tripartite characterization of moral realism, I take non-naturalist realist views to feature the following properties: (i*) moral facts are mind-independent, nonrelative, and causally inert; (ii*) we have beliefs that aim at representing them truly; (iii*) we know some of these truths via a priori rational intuition (where a priori rational intuition is not reducible to something clearly naturalistic, such as trained expert judgment of facts that also are empirically accessible). My dilemma takes aim at non-naturalist views that feature (i*)–(iii*). However, moderate metaethical views, whether realist, constructivist, or expressivist, are unthreatened by my dilemma, for they connect our moral judgments more closely to our attitudes or to properties with which we stand in close contact.Footnote ³ This is why I call them moderate.

Perhaps not everyone who considers themselves a non-naturalist accepts (i*)–(iii*). For example, one might hold that moral properties are causally inert and hold a coherentist epistemology without intuitions, or one might do without moral knowledge tout court. However, (i*)–(iii*) together represent a paradigmatic non-naturalist position. It is compatible with both robust non-naturalism, according to which moral properties are substantive and irreducible to natural properties yet causally inert (e.g., Enoch Reference Enoch2011; Huemer, Reference Huemer2004, Reference Huemer2016; Moore Reference Moore1903; Shafer-Landau Reference Shafer-Landau2003), and relaxed or quietist non-naturalist realism, which entails that moral properties do not carry ontological weight (e.g., Dworkin Reference Dworkin2011; Parfit Reference Parfit2011; Scanlon Reference Scanlon1998, Reference Scanlon2014; Skorupski Reference Skorupski2010).

Importantly, (i*)–(iii*) are also commitments of Huemer's. More specifically, Huemer thinks there are objective, mind-independent, and causally inert moral truths that we—or at least experts, who may disperse their knowledge socially—originally arrive at via synthetic a priori intuition (Huemer Reference Huemer2016: 1985–88; 2000; 20006; cf. Huemer Reference Huemer2004). While he does not want to commit himself to a theory of how such intuition works, he suggests that it should not be treated as an empiricist source of knowledge (Huemer Reference Huemer2016: 1986–87). But he does think that we have arrived at some of the moral truths, as we soon shall see. And it is Huemer's view that shall be my primary foil when I develop my dilemma below.

There is a further complication, however. This is because there are two kinds of moderate metaethical views: those that take there to be nonrelative moral truths that we know and those that do not. I call the former group of views nonradically moderate, as they allow for a nonskeptical vindication of ordinary moral discourse, and the latter group radically moderate, as they appear unable to do so. Nonradically moderate views include most types of naturalist moral realism, objectivist constructivism (e.g., Kantianism), and quasi-realist expressivism. These views all purport to explain objective moral truths and knowledge, albeit in different ways. While correspondence accounts of truth and cognitivism about moral judgments fit realism well, Kantians may perhaps appeal to transcendental arguments to generate objectivity, whereas quasi-realists tend to prefer deflationary accounts of truth together with expressivism about moral judgments. Radically moderate views are, instead, those that deny that there are nonrelative moral truths or that we have moral knowledge: for example, extreme forms of metaethical relativism, error theories, and some early forms of expressivism such as emotivism.

The contingentist dilemma supports nonradically moderate views over non-naturalism, but not radically moderate ones. This is because radically moderate views cannot explain moral progress. The contingentist dilemma is premised on moral progress, but there is no progress to be made if there is no moral knowledge or no nonrelative moral truths on which to converge. However, I assume that the existence of at least some moral progress is a datum. I shall therefore ignore radically moderate views for now. (Defenders of such views may want to read the rest of this paper hypothetically.)

To summarize, it may be helpful to contrast my terminology with Huemer's. For him, antirealist views fail to explain moral progress. These are views according to which there is no moral knowledge or are no objective moral truths to know (Huemer Reference Huemer2016: 1983). These are the views I have called radical moderate metaethical ones. I am happy to agree with him here: any kind of realism is likely to win out in contrast with these views. However, Huemer does not discuss alternatives to the non-naturalist realism he defends or to the antirealist or radical moderate views he criticizes. He does not mention quasi-realist expressivism or objectivist constructivism at all, and he takes naturalist realism to be beyond the scope of his paper (Huemer Reference Huemer2016: 1986, fn. 11). If my argument works, however, it will lend support to the views I have called nonradically moderate. Non-naturalist moral realism explains moral progress worse than they do.

3. The Contingentist Dilemma

Now we know our -isms and aims. Here is the new dilemma:

• (1) Our moral attitudes are to a significant extent shaped by contingent influences.

• (2) Either there is a relation between these contingent influences on our moral attitudes and the non-naturalist moral facts, or there is not.

• (3) If there is no relation, the contingent influences are extremely likely to have put our moral atittudes off track from the non-naturalist moral facts.

• (4) If there is a relation, these contingent influences tend to make us know the non-naturalist moral facts, but this is unlikely to be the case on any good explanation of moral progress.

• - - -

• (C1) Either way, it turns out that we do not know the non-naturalist moral facts.

• (6) If we do not know the non-naturalist moral facts, moral non-naturalism is false.

• - - -

• (C2) Moral nonnaturalism is false.

The contingentist dilemma has clear similarities to Street's dilemma. The move from (C1) to (C2) by way of (6) is identical, and the rest of the argument has roughly the same shape: the disjunction introduced in premise (2), premise (3), and the point about explanatory unacceptability in (4) are Street-inspired

What differs is the focus on generalized contingency rather than evolution and, hence, especially premises (1) and (4). These are the controversial premises, and I shall discuss them at length in sections 4 and 5 below. However, I take Street's defenses of her versions of premises (3) and (6)—the other premises that need defense—to be fundamentally plausible, so I shall start off by briefly explaining how analogous defenses support my versions of them.

As far as premise (3) goes, Street thinks that if there is no relation between the evolutionary influences on our moral attitudes and the mind-independent moral facts, evolution will almost certainly have led us off track. My version of premise (3) says that contingent influences are likely to have pushed our moral beliefs off track from the moral truths, non-naturalistically construed. Street's argument for her conclusion is that evolution does not select for mind-independent moral truths. Similarly, contingent influences—by which I mean unreliable sociocultural or psychological influences, on top of evolutionary ones—will not lead us toward non-naturalistic moral truths if there is no independent relation between the contingent influences and those truths. Hence, I think Street's core point can be generalized, with the caveat that it holds between influences on our attitudes and non-naturalist moral truths rather than all mind-independent moral truths, as per the last section.

As far as premise (6) goes, Street mostly relies on it implicitly. But the premise is widely held, not least among non-naturalists. I have therefore included it in my characterization of moral realism as Street understands it above as well as in my own characterization of moral non-naturalism. It is, however, logically possible to accept non-naturalist-style metaphysics and a rationalist intuitionist epistemology without thinking that we have moral knowledge. But many would count it as a significant loss to give up on moral knowledge, and so it will be fair to run with the premise for now. (However, those who disagree may want to read the rest of the paper hypothetically.)

4. In Defense of Premise (1)

Premise (1) says that our moral beliefs are to a significant extent shaped by contingent influences. I shall argue that this is so primarily because our attitudes are shaped by sociocultural and psychological factors. There may also be evolutionary forces at work, but these are put in the background relative to the former.

What does ‘contingent’ mean here? I mean to count a broader set of factors than evolutionary ones as contingent. I stipulate that a factor or influence on our moral attitudes is contingent if it may affect our attitudes in a way that does not dispose them to be reliably connected to mind-independent and causally inert moral truths. This does not mean that noncontingent factors necessarily will lead to true beliefs—they just do it reliably—or that contingent factors cannot lead to truths—they just do it unreliably.

There are two caveats. First, this use of ‘contingency’ should not be conflated with other uses of the term in metaethics, such as denials of the claim that moral properties are necessary properties. Rather, contingency is a property of potential factors that may affect our normative attitudes. Second, contingent factors are also likely to affect our attitudes about descriptive facts. But there their impact is mitigated by our ability to double-check what is going on using ordinary empirical means: for a made-up folkloristic example, perhaps someone has been taught that the howling of the wind at midnight is the sound of a ghost, but further investigation is likely to show that it is just the sound of the wind. We could double-check by going for a second look, infer that the sound is just the howling of the wind from general background knowledge, or recreate the wind in a controlled experimental setting and hear whether it makes a similar sound. As such, I want to emphasize how contingent factors can affect our judgments about the noncausal, mind-independent moral truths that non-naturalist realists posit.

Which contingent factors are there? Evolution aside, Street, like many others, allows that there can be many contingent factors that influence our beliefs. She writes:

To systematize these ideas, we may distinguish between three types of contingent influences on our attitudes. They are rough and a first stab, but they are nevertheless a helpful tool for a first analysis. First, then, there are evolutionary influences, whether these stem from natural selection or not. Second, there are the sociocultural ones, or what Street calls ‘social, cultural, historical’ influences. Third, there are psychological factors, including various biases, emotions, and mental associations, as well as rational reflection itself—for not even rational reflection must be reliably connected to noncausal mind-independent moral truths.

Why? Sometimes we reason about what to do based on values or goals that are unlikely to be connected to such truths. Whether someone who is working out should go swimming or jogging does not seem like a moral issue, just a prudential one. And even if we, quite plausibly, believe that ethical egoism is a false theory, we can recognize that someone who believes that ethical egoism is true but believes that donating to charity at the detriment of their own interests is good could make their views more coherent by dropping the latter belief.

Contingency is now characterized. We may then wonder which influences, contingent or not, there are on our attitudes. Street's evolutionary forces are a first possibility. However, Street does not say anything about why we should take them to be more prominent than sociocultural or psychological forces when she sets up her argument. Of course, she argues that evolutionary forces are more prominent than mind-independent moral truths, but her arguments do not touch on whether evolutionary or other contingent forces are the most explanatorily prominent ones.

This is a weakness in Street's argument, and Huemer challenges Street on this point by bringing moral progress into the picture. While Huemer, quite rightly, agrees with Street that evolution and culture both may be part of the full explanation of our moral attitudes (Huemer Reference Huemer2016: 1986), moral progress presents a problem for Street's emphasis on evolutionary forces as an explanation of our attitudes. For if we have progressed morally over time toward some particular values, the progress can be taken to depend on something else than evolutionary factors.

Why is that? In general, I take the phenomenon of moral progress to be a shorthand way to discuss how things, in general, become morally better: for example, slavery gets abolished, sexism decreases, or political institutions become appropriately democratic. But regardless of what drives moral progress or whatever else moral progress involves, I agree with Huemer that a necessary and substantial feature of it on any plausible account is changing moral attitudes: moral progress involves them improving over time. That is likely to be both constitutive of various improvements (for example, fewer sexist beliefs means less sexism) and an important source of further change (for example, few nondemocratic countries will become democracies if no one believes democracy is better than dictatorship).

How can we accommodate changes in moral attitudes in an account of moral progress? Almost certainly, changes in moral attitudes are going to involve contingent sociocultural and psychological factors. Among candidate sociocultural factors, Huemer (Reference Huemer2016) especially emphasizes social learning. Sometimes moral experts make genuine discoveries. But most of our moral learning, he thinks, stems from the transmission of moral knowledge, whether from experts to the folk, from the folk to the folk, and—presumably—sometimes even from the folk to experts. I concur; this is very plausibly a feature of the explanation of moral progress. But information transmission is contingent, for falsehoods are also easily transmitted.

Moreover, it also seems overwhelmingly likely that a significant amount of our moral reasoning is contingent. It is often based on whatever values we hold, independently of whether these are connected to moral truths. Following Campbell and Kumar (Reference Campbell and Kumar2012), Hopster (Reference Hopster2020), for example, suggests that moral reasoning consists of rooting out inconsistencies among our attitudes. This could be so even if we hold the (in my view obviously flawed) position of ethical egoism, as in the example above.

There are further sociocultural and psychological possibilities. First, there is intergroup association: people adopt the beliefs of their in-groups and avoid those of out-groups. Second, there are moral intuitions that are likely to be at least sometimes contingent by not reflecting any moral truths but rather dominant social attitudes, prejudices, snapshot (non-morally) motivated judgments, and so on, and which nevertheless come to influence our moral beliefs (Haidt Reference Haidt2012). Third, there might be indirect influences on our moral expectations that stem from more or less sincere attempts to express moral views in public (Westra Reference Westra2021).

All the factors just mentioned are likely to be involved in the formation and transmission of moral attitudes between people. I shall not endeavor to argue which one of them is the most prominent: what matters is that they all are plausible candidate mechanisms for explaining attitudinal change. They are the sort of things that are likely to bring it about.

Of course, none of these factors rule out influences on our moral beliefs that are evolutionary or noncontingent. But it seems highly likely that at least some of the sociocultural and psychological factors just mentioned ought to feature in any plausible account of the development of our moral views insofar as these have made progress over time. They play an important role in the explanation of our current moral attitudes.

In fact, these factors seem to do so to the extent that purely evolutionary forces cannot be the most central relevant explanatory factors when it comes to the details of our current attitudes. Our moral attitudes have changed recently and rapidly, and it is hard to think that is all due to biological factors. There has not been enough time for evolution to shift them this much this quickly, and there is no reason to think attitudes generating or constituting progress have been more evolutionarily successful than others over the last fifty years or so (Huemer Reference Huemer2016: 1995). In addition to Huemer's points, we may add that many progressing moral attitudes are affected by developments (whether social, theoretical, technological, or environmental) that biology does not say much about. It is hard to take a stance on what to think about social democracy or AI ethics without the concepts of ‘socialism’ and ‘democracy’ and ‘AI’. As such, contingent sociocultural factors are likely to have played a key role in the dispersal of progressing moral attitudes; indeed, such factors may sometimes be so strong that they could cancel whatever impact evolution has on our attitudes.Footnote ⁴

Street might seem able to agree with this: she claims only that broad evaluative dispositions are evolutionarily selected for, so sociocultural forces are not ruled out by her account. She seems committed to what we may call a broad adaptive link account, which does not tie specific judgments but rather basic evaluative dispositions that can give rise to many different judgments to our evolution. A broad adaptive link account can, therefore, accommodate a significant amount of moral change or progress. This would be different from a narrow adaptive link account, which would tie evolution to specific evaluative judgments, hence making rapid and recent moral change hard to explain (as per the above). For this reason, Hopster (Reference Hopster2020) claims that Huemer's argument about the role of moral progress in response to the Darwinian dilemma has little impact on it.

However, even a broad adaptive link account is misleading: it is too broad. It leaves an explanatory gap where other explanatory factors do the most important work to explain our current attitudes. Recent and rapid moral progress indicates that our moral attitudes are highly variable, and we need something that explains why they have changed into our current, fine-grained attitudes—about things like social democracy and AI ethics and potentially also about things that are evolutionarily unhelpful—rather than into broad dispositions that could give rise to many different judgments. We need this because the broad adaptive link account has its explanatory factors too far down the causal chain from our current attitudes to say why we have our current attitudes. While we very plausibly could have some broad evolved dispositions—for example, toward cooperation—explaining current attitudes with them would not provide much information about why we have these specific attitudes rather than others. This is comparable to the way in which ‘We have our current moral attitudes because life once appeared on earth’ or ‘We have our current moral attitudes because the Big Bang was the start of the universe’ are uninformative explanations. They plausibly pick out parts of the causal chains that lead to our current attitudes, but they leave a gap to be filled in.

The upshot is that the broad adaptive link account seems insufficiently detailed to explain our current attitudes. It says nothing by itself about shifts, and it is not fine-grained enough. While it is quite possible that evolutionary factors may have some background effect on our moral judgments, whatever explanatory force they have leaves too much of a gap in the explanation of our current attitudes, so they are comparatively speaking backgrounded by sociocultural and psychological factors.

Huemer does, however, challenge those who want to emphasize sociocultural factors in their explanation of moral change to too great an extent. He argues:

The core idea here is that the systematicity of moral progress indicates that it cannot just be contingent sociocultural factors that shift our attitudes. But I am happy to agree with Huemer. Sociocultural factors need not be the grounds of a debunking explanation, in general, but they may form part of a full explanation of moral progress, while the full explanation also may incorporate evolutionary and psychological factors—and moral truths. This is even how social learning features in Huemer's theory.

However, this point of agreement is independent of whether we should go with non-naturalist realist or nonradical moderate metaethical views. As mentioned in section 2, Huemer contrasts non-naturalist realism with radical moderate metaethical views (which he calls antirealist). But the nonradical metaethical moderate can accept all of Huemer's points and simultaneously claim that moral facts, truths, or properties are part of the explanation of moral progress, regardless of which interpretation of these she may prefer (other than non-naturalism). For example, the fact that slavery is bad may explain why slavery has been abolished even if its badness is grounded in our attitudes. Hence, the mere fact that our moral attitudes often have contingent influences, whether sociocultural ones or others, does not rule out moral progress.

Accordingly, it is plausible that contingent sociocultural and psychological factors are likely to have influenced our moral outlooks to a significant extent. This does, comparatively speaking, downplay evolutionary influences and the influence of reason (understood noncontingently) on our attitudes. However, it does not rule out saying that such factors may have some influence on our moral attitudes, though they are likely to be for present purposes less explanatorily relevant than contingent sociocultural and psychological factors. Hence, premise (1) is defended.

5. In Defense of Premise (4)

Premise (4) reads: ‘If there is a relation [between the contingent influences on our moral attitudes and moral reality, non-naturalistically construed], these contingent influences tend to make us know the non-naturalist moral facts, but this is unlikely to be the case on any good explanation of moral progress’. Why? Non-naturalists need to explain how moral progress featuring contingent factors leads to moral knowledge. But it is unlikely that any good theory about the emergence of our current moral attitudes, construed as featuring moral knowledge after a process of moral progress, will feature causally inefficacious moral facts known through non-naturalist-style intuition. This presents a significant challenge for non-naturalism. It is likely to provide a less virtuous explanation of moral progress than explanations that fit better with nonradical moderate metaethical views.

I have so far been sympathetic to Huemer's criticism of Street, but he has also sketched a positive explanation of moral progress in terms of experts intuiting the non-naturalist moral facts that later get dispersed through societies. Here, I want to develop a challenge to it—that also ranges far beyond Huemer's own view. Now, I do not doubt moral progress. I shall assume that it has occurred, and I am happy to follow Huemer in assuming that a significant part of it involves converging on the right values—even the values he has identified.

But to explain moral progress in terms of changing attitudes is to explain a social phenomenon: namely, how moral attitudes have changed to become better than they once were. Explaining this kind of descriptive yet normatively loaded phenomenon is ordinarily something the social sciences deal in. It is analogous to explaining democratization, good governance, economic development, or what have you. This means that explaining moral progress calls for a broadly speaking social scientific explanation and therefore an explanation that is responsive to the ordinary virtues of scientific theories, such as explanatory power and simplicity, when it is contrasted with other plausible explanatory hypotheses. Because of this similarity, anyone saying that we should invoke extrascientific entities—whether theological, philosophical, or other—or use some nonstandard scientific method in such an explanation faces the challenge of justifying such methodological choices.

Assuming then that we may make use of ordinary scientific standards of theory choice, we can form many plausible hypotheses that are likely to fit the data. They may do so in whole or in part depending on their scope—and their scope may reasonably vary, for whether a unified theory of all moral progress is correct is an empirical question. Perhaps we will then be able to form a unified general theory such as: ‘Increased socioeconomic standards lead to increased convergence on liberal values’. Or perhaps we need a more complex theory with several subhypotheses, such as: ‘Increased economic integration between countries leads to increased appreciation of others, which leads to increased peacefulness, and increased access to interaction with people from different walks of life leads to an increased appreciation of human dignity’. Or we may explicitly invoke (naturalist-friendly) moral properties, such as ‘Hitler's badness leads to increasingly dispersed knowledge of his badness, which encourages appreciation of human dignity’. It does not matter which theory we go with; they are all plausible candidates for explaining moral progress. And they are compatible with nonradical moderate metaethical views; they do not involve non-natural moral truths known through rational intuition.

What I shall argue is that, in contrast with theories like these, Huemer's explanation conflicts with many key virtues of any good scientific explanation of moral progress. When comparison with a more detailed theory (rather than between Huemer's view and any possible plausible alternative) is called for, I shall use the first disjunct of the second theory above—namely, explaining increased peacefulness between countries in terms of people thinking better of each other because countries become increasingly economically integrated—as a point of comparison with Huemer's view. But the points will apply in comparisons between Huemer's theory and any plausible alternative theory, so the reader is free to change to some other example.

My suggestion, then, is that the lack of virtues of Huemer's view makes it implausible. But the poverty of his explanation generalizes: it is problematic because of its appeal to non-naturalist-style a priori moral intuition and causally inert moral properties. I therefore hypothesize that any theory that appeals to non-naturalist-style a priori knowledge of causally inert moral properties will be a bad theory of moral progress. Yet non-naturalists are committed to doing that, for they must address how we have progressed to know the moral facts. While it has yet to be determined what the best theory of moral progress is, non-naturalist explanations therefore appear to be poor theoretical alternatives in contrast with theories that are compatible with nonradical moderate metaethical views, regardless of which one ultimately is best.

Accordingly, I will not argue in favor of any specific explanation of moral progress compatible with such views here. That is too ambitious for now. But there is nevertheless a strong prima facie case against non-naturalist-style explanations of moral progress based on considerations of theoretical virtue. While there is significant controversy about how the theoretical virtues should be understood, I shall pick out a considerable number of key virtues that any theory ought to be able to handle well, and argue that Huemer's non-naturalist explanation does poorly on many of them. That is enough for an argument against him and, I suspect, against non-naturalism more broadly.

Now to the explanations. Huemer thinks broadly liberal values have become increasingly endorsed over time.

The liberalism that has become endorsed need not be any particular version of liberalism but only this: ‘liberalism (1) recognizes the moral equality of persons, (2) promotes respect for the dignity of the individual, and (3) opposes gratuitous coercion and violence’ (Huemer Reference Huemer2016: 1987). Points (1)–(3), he thinks, hang together systematically. One may quibble about the details—for example, are these the exact values that have become endorsed, and should they be called ‘liberal’, ‘humane’, or ‘progressive’?—but Huemer's overarching point is plausible. The values in (1)–(3) often steer our judgments, either in general or in applications—including on topics as different and fine-grained as social democracy and AI ethics—or gain support from our judgments about various topics or cases themselves.

Huemer thinks we have converged on these values because we have made moral progress. To explain that, he develops an account of moral progress according to which we have intuited moral truths even though these truths are causally inefficacious. On this account, the development of moral knowledge over time follows the same pattern as the development of scientific knowledge. In less developed societies, there is widespread disagreement. But both scientific and moral beliefs develop over time to become more subtle, complex, and technical. We should expect widespread convergence in due time because of how the beliefs are transmitted through our cultures. While convergence may take centuries or millennia to achieve, we do observe that achievement occurring in reality.

How can we trust the beliefs we have converged on through cultural transmission? Here, too, Huemer makes an analogy with scientific beliefs. Because science is generally reliable, we can trust many culturally transmitted scientific beliefs. Something similar holds in ethics. Huemer assumes (Reference Huemer2016: 2003–4) the following:

1. (a) Human beings possess a capacity for a priori knowledge.

2. (b) Human beings can gain a priori knowledge of morality.

3. (c) Human beings have various nonrational influences of our beliefs, including evolutionarily or culturally caused emotions and desires.

4. (d) Cultural influences make it hard for individuals to believe something radically at odds with the cultural mainstream, but it is relatively easy to believe something that is moderately at odds with it.

5. (e) Individuals differ in their sensitivity to moral truth as well as in their susceptibility to nonrational influences on their moral beliefs.

In less developed societies, we will be systematically morally wrong, but in virtue of our ability to intuit moral truths, we are able progress from there. Often change happens in approximations and incrementally, but nevertheless there is a ‘systematic tendency’ toward moral development (Huemer Reference Huemer2016: 2005).

The mechanism behind the tendency is one where reformers are more rational than ordinary members of societies and also tend to reach positions of power, making their views influential enough to generate broader societal reforms. Once some progress has been made, more progress follows by additional iterations of the reform process, generating morally better societies over time. While moral truths do not contribute to this process causally, there is still a lawlike generalization about how our increasing understanding of liberal morality may bring the process about; thus, there is a sense in which the moral truths we have intuited explain our current practices.

Huemer's view has some virtues. He does not make ad hoc assumptions (Huemer Reference Huemer2016: 2003–4), he unifies explanations of scientific and moral progress, and his view has intertheoretical support because he appeals to assumptions that non-naturalists are likely to endorse anyway. Nevertheless, his view does poorly on other virtues. These are generally considered key ones, and any theory that fails to make good on them is likely to be worse off than other theories. The relevant virtues are these:

• • Qualitative Parsimony

One familiar virtue is qualitative parsimony (cf. Hopster Reference Hopster2020; Street Reference Street2006). This means that a theory should appeal to as few entities or categories as possible in otherwise good explanations.

Huemer's view does not do this. It features a priori knowledge of moral truths gained through non-naturalist-style rational intuition and causally inert moral properties. But we could, potentially, do without both. Most metaethical views and all more typical scientific explanations do.

• • Explanatory Power

I follow Ylikoski and Kuorikoski (Reference Ylikoski and Kuorikoski2010) in interpreting explanatory power. Hence, by ‘explanatory power’, I mean the ability of a theory to explain relevant data well, for some relevant interpretation of ‘well’. For example, the theory increases our understanding, is insensitive to changes in background conditions, has a precise explanandum, increases the factuality of our beliefs, integrates new information with old, is cognitively salient, or something else along those lines (or some conjunction of some subset of those properties).

Huemer-style non-naturalist realist explanations of moral progress could, potentially, help us with some of these properties. While Huemer has a fairly clear explanandum—the moral progress toward liberal values—what one thinks about questions of factuality, information integration, and cognitive salience is likely to turn on whether one previously has been sympathetic to non-naturalism. A non-naturalist will likely think that Huemer's explanation increases the factuality of our beliefs, integrates new information with old, and is cognitively salient. But others will disagree if they have other background theoretical commitments.

Nevertheless, I want to emphasize problems with two of Ylikoski and Kuorikoski's criteria. First, it is highly doubtful whether Huemer's explanation increases our understanding. His explanation relies on non-naturalist-style causally inert moral properties and rational intuition. But, as we learn inter alia from Mackie's famous queerness argument (Reference Mackie1977) and the ensuing debate, making sense of non-naturalist-style moral properties and rational intuition has proven notoriously elusive. Explaining some phenomenon with exogenous variables we do not understand well is not going to increase our understanding of the phenomenon much.

Second, non-naturalist realist views appear to be highly sensitive to changes in background conditions, such as whether there is non-naturalist-style moral knowledge. If it turns out that the experts have gone systematically wrong, their explanation falls apart. Thus, Huemer's explanation does not seem to be particularly robust.

The two major issues for Huemer when it comes to explanatory power, then, involve understanding and applicability across a range of background conditions. As far as understanding goes, our understanding of moral progress will increase more if we can specify the mechanisms behind it in terms other than non-naturalist ones. This is because most such explanations are likely to be more conducive to understanding than non-naturalist ones, for they are not based on assumptions as mysterious as those of non-naturalist views: the details can be filled in with familiar facts. Perhaps economic integration generates increased peacefulness by itself or by way of another variable such as people's improving attitudes stemming from economic integration.

As far as robustness goes, an explanation of improving moral beliefs in terms of beliefs that can be formed by many people in a population—such as in countries that are becoming economically integrated—is likely to be better than an explanation that requires beliefs that only are formed by a few experts. This is because there is a greater pool of potential believers.

• • Conservativeness

By ‘conservativeness’, I mean the extent to which a theory fits with and is based on our previously accepted theories. This is similar to the idea of information integration mentioned in the discussion of explanatory power. Hence, there will likely be different views about whether Huemer's non-naturalist realism is conservative, depending on whether one is sympathetic to its assumptions from the start.

However, I argued above that we should start from typical scientific assumptions in an explanation of moral progress, in which case non-naturalism has a problem here. This is because scientific explanations of social developments do not ordinarily feature non-naturalist moral properties or rational intuition (cf. Hopster Reference Hopster2020). Rather, they feature biological facts, sociocultural facts, psychological facts, institutional facts, scientific laws, and so on. Non-naturalist moral properties and intuitions are rather alien to biological, psychological, or social scientific explanations. This means that all kinds of scientific and naturalistic views are likely to do better than Huemer's here too.

• • Fruitfulness

By ‘fruitfulness’, I mean the extent to which a theory can be used to make new theoretical developments. I do not think Huemer's view does well here either. We are not likely to be able to formulate scientific hypotheses about why certain moral beliefs have been formed at some rather than other points in history or in some rather than other cultural contexts on his view. For why and when should we believe that some expert has or has not grasped a truth? Nothing seems to forbid the possibility that both internally, from the expert's perspective, and externally, from an observer's perspective, the grasping of a truth and a falsehood can look exactly the same. Accordingly, Huemer's view does not seem very theoretically fruitful.

However, if we instead formulate our explanatory hypotheses using less contestable facts, hypotheses involving predictions based on the presence of specific explanatory factors can be formulated more easily. For example, if increased economic integration between countries generates increasingly positive attitudes that in turn increase peacefulness, we can predict that peacefulness will increase when there is increased economic integration.

• • Testability

By ‘testability’, I mean the extent to which a theory can be used to formulate hypotheses that make it possible to confirm or disconfirm with empirical data. It matters that the data here are empirical: they can be observational or experimental, come from lab or natural experiments, or something else. But insofar as we want a scientifically virtuous theory, it should be confirmable or disconfirmable on empirical rather than just theoretical or intuitive grounds.

Huemer's view is not likely to be testable (in any interesting sense, at least). Admittedly, it will be confirmed if we have converged on liberal values and disconfirmed if we have not. We can assume that we have. But that datum does not lend much support to the theory. Many other explanations of convergence are possible, and the causal inefficacy of Huemer-style moral properties entails that their existence is not in the same ballpark as causally efficacious ones that might be put to causal tests. Non-naturalist moral properties are either there or not whatever happens: murder will be wrong even if no one is ever murdered.

The same issue appears regarding the discovery of these properties. When do we know when we have discovered moral properties or made false ‘discoveries’? Huemer does not tell us how to measure this. We do not have any predictions about why or when any particular moral truths would be intuited at any point in time. By contrast, as mentioned in the discussion of fruitfulness, naturalistic theories, such as the one that suggests that economic integration drives peacefulness by way of positive attitudes, easily provide predictions. Hence, they are also easy to test.

Summing up, Huemer's explanation does poorly when it comes to several key theoretical virtues: qualitative parsimony, explanatory power, conservativeness, fruitfulness, and testability. Other views, including broadly naturalist ones, do better. Here, then, is the generalization hypothesis: explanations of moral progress that do not appeal to causally inert moral properties or non-naturalist-style intuitions are likely to do better than those that do. This is because the same type of considerations as those I have presented against Huemer will count against other explanations that invoke these aspects of non-naturalism.

Implicitly, this discussion has also generated replies to some potential objections. Some have suggested that charging non-naturalism with not being scientifically virtuous just rehearses stock arguments against it (e.g., claiming that it lacks parsimony). But what I am concerned with is the (lack of) virtues it has as a part of an explanation of moral progress, not its (lack of) virtues as a theory about the constituents of the world.

Others have suggested that this setup might overreach and indicate that no metaethical theory will end up being part of a virtuous explanation of moral progress. But the proof will be in the pudding. I am not committed to any particular metaethical or social scientific theory, and whether the best explanation ends up involving moral properties, quasi-realist attitudes, social scientific institutions, or something else is up for grabs. I have rather argued against a theory that appears to be poor, whatever the best one will involve.

As such, premise (4) is defended. (C1) follows. And as premise (6) has been assumed, (C2) follows. Thus, moral non-naturalism is likely to be false because it cannot explain the moral knowledge that non-naturalists assume we have. This is because of the contingentist dilemma.

6. Recap and Third Factors

I have defended the contingentist dilemma for moral non-naturalists. It is based on moral progress and the contingency of our moral attitudes suggested by it. If we assume non-naturalism, either there is a relation between contingent influences on our moral attitudes and non-naturalist moral facts, or there is not. If there is no relation, the contingent influences will almost certainly have put our moral attitudes off track from the non-naturalist moral facts. If there is a relation, these contingent influences would have to tend to make us know the non-naturalist moral facts—but this is scientifically unacceptable, as non-naturalist style explanations of our moral knowledge via moral progress are likely to be worse than explanations that do not feature non-naturalist metaphysical and epistemological assumptions. Thus, we are not likely to know the moral facts on moral non-naturalism. But non-naturalists assume we do know them. Consequently, non-naturalism is false. However, as the argument presumes moral progress, it supports nonradical moderate metaethical theories rather than radical ones.

So far, I have not discussed the worries that often are presented for Street's Darwinian dilemma beyond Huemer's moral progress worry (which I also endorsed). The literature on Street's dilemma is, however, vast and developing, so it is impossible to deal with it in full depth here. But I want to discuss the most prominent objection to her view: third-factor explanations (for defenders, see e.g. Behrends Reference Behrends2013; Enoch Reference Enoch2010; Reference Enoch2011; Moon Reference Moon2017; Skarsaune Reference Skarsaune2011; Talbott Reference Talbott2015; Wielenberg Reference Wielenberg2014; for critics, see e.g. Dyke Reference Dyke2020; Klenk Reference Klenk2017; Korman and Locke Reference Korman, Locke and Shafer-Landau2020). I shall argue that the contingentist dilemma remains an issue for non-naturalism even if third-factor explanations are plausible responses to Street. This means that the contingentist dilemma does not just improve on Street's dilemma regarding moral progress: it also improves on it dialectically by presenting a challenge for non-naturalism that handles this key objection.

Third-factor explanations are problems for Street's version of premise (3). The core idea behind them is that there is a third factor that explains both why our current moral attitudes have evolved and why they are likely to be true. I lack the space to address all possible third factor explanations here, but my dilemma avoids the most prominent ones. Perhaps the most prominent of all is David Enoch's (cf. Enoch Reference Enoch2010, Reference Enoch2011). He suggests that the relevant factor is the value of survival: evolution makes us form beliefs conducive to it, and as these are in line with the mind-independent moral truth that it is of value, we are likely to form broadly accurate moral beliefs.

However, this explanation lacks bite in response to my version of premise (3). Street argues that evolutionary forces are likely to have taken us off track from the non-naturalist moral facts, but regardless of what we make of Enoch's proposal, it is plausible that there could be a third factor that links up evolutionary success and morality. This is because humanity has evolved as one single species, and there might be something about that species that is such that its attitudes are correlated with morality. It might, for example, value survival.

I have, however, emphasized sociocultural and psychological rather than evolutionary factors. Contra Enoch's response to Street, it is extremely implausible to think that there could be a third factor linking these kinds of contingency with non-naturalist moral facts. Beliefs are heterogeneous; we can and do think pretty much anything is true, whereas our evolution is ‘fixed’ in the sense that we are just one species with one shared evolutionary history. Hence, beliefs affected by contingent sociocultural and psychological factors are much more unlikely to have much to do with causally inert non-naturalist moral truths than evolutionarily caused beliefs do.

This point also generates a reply to so-called evolutionary byproduct third-factor responses. According to these, evolution has given us a general capacity to form true judgments. The ability to form true moral beliefs is a byproduct of that capacity (cf. Deem Reference Deem2016). This is another third factor explanation: a general capacity to form true beliefs has led both to our current true beliefs and to an ability to track moral truths, even though it is a byproduct of this faculty that we get morality right.

I think this is a very strong response to Street's evolutionary contingency, but it has no bearing on my types of contingency. Nothing about sociocultural and psychological contingency indicates that there is a link between moral judgments influenced by these types of contingency and non-natural moral facts: pretty much any conceivable moral view might become fashionable, so our judgments can go wrong in pretty much any way. We still need a story about why standard moral reasoning would be related to the moral truths, non-naturalistically construed, to see why there would be a relation between our beliefs and those kinds of truths when we consider contingencies other than evolution. But none has been given. Hence, the byproduct response does not solve the contingentist dilemma.