J. L. Mackie (Reference Mackie1977, 37) claimed, “the actual variations in the moral codes are more readily explained by the hypothesis that they reflect ways of life than by the hypothesis that they express perceptions, most of them seriously inadequate and badly distorted, of objective values.” For Mackie, what we take to be morally right or wrong and good or bad is best understood as a function of how we live our lives. We do not apprehend objective moral facts or values, and our moral codes are not based on perceiving mind-independent or constructed moral facts. Rather, he thought, different moral codes result from our divergent ways of life and social practices (see Bagnoli Reference Bagnoli2015; Street Reference Street2010). Theorists such as John Doris, Alexandra Plakias, and Stephen Stich have since sought to defend the basic view that Mackie articulated (Doris and Stich Reference Doris John, Stich, Jackson and Smith2005; Doris and Plakias Reference Doris John, Plakias and Sinnott-Armstrong2008; Machery, Kelly, and Stich Reference Machery, Kelly and Stich2005, 830). These theorists have developed rigorous, empirically backed arguments for the claim that the prevalence and depth of actual moral disagreement put moral realism in doubt. As I will understand it, moral realism has four component views. The realist believes that: (i) moral claims can be true or false (cognitivism holds), (ii) at least one moral claim is true, (iii) moral truths are “objective” in that they are “independent of human perceptions, attitudes or practices[,]” and, many realists add, (iv) human beings know some of these truths (Hassan Reference Hassan2019, 146).Footnote 1 Moral realism so construed is, in short, the view that “moral truths are underwritten by mind-independent moral properties or facts and that these properties or facts are epistemically accessible to moral agents” (Hopster Reference Hopster2020, 1256). To what extent, then, do the empirics of actual disagreement call realism into question?Footnote 2
Empirically minded antirealists target convergentism in particular. This is the view that “rational individuals who were fully informed of the nonmoral facts would not hold incompatible moral views” (McGrath Reference McGrath2010, 59).Footnote 3 Simon Fitzpatrick (Reference Fitzpatrick2014, 162) says why moral realism might require convergence: If moral truths are epistemically accessible, per condition (iv) above, this “would seem to imply a gradual convergence in moral views over time, as individuals and communities engage in inquiry.”Footnote 4 While some moral realists deny that realism requires convergentism, my focus in this essay will be convergentist realism (henceforth, sometimes just realism), which is “the industry standard in contemporary moral realism” (McGrath Reference McGrath2010, 59). Focusing on Doris and Plakias’s (Reference Doris John, Plakias and Sinnott-Armstrong2008, 313 passim) prominent discussion, I shall suppose, arguendo, that realism requires convergentism. This assumption will let me address directly the Doris-Plakias challenge to realism: specifically the claim, detailed later, that actual and not just ideal disagreement poses a threat to moral realism. Convergentist antirealists of this ilk identify empirical cases of moral disagreement and claim that no “defusing explanations” show it to be compatible with realism.
I will argue that this crucial antirealist move overlooks important defusing explanations, some of which put in doubt whether any empirical study of moral disagreement could undermine convergentism. My defense of realism starts by showing that scenarios such as the prominent “Magistrate and the Mob” (MM) case betray cognitive defects in subjects, such as partiality, that we would not find in ideal agents but should expect to find if antirealism is true. After this, I develop and defend three defusing explanations for why we should expect considerable cross-cultural disagreement on moral norms even if convergentist realism is true. These explanations concern (i) individual and (ii) group moral development, plus (iii) the moral models that agents use. While developing my defense of moral realism, I aim for comprehensive engagement with responses to arguments by Doris, Plakias, and others that have been dispersed across several related articles.
My overall argument carries important implications for how we should understand recent challenges to convergentist realism. Philosophers have recently argued that realist accounts of historical convergence in moral views (e.g., Huemer Reference Huemer2016) have less evidential value than the realist might hope for. Cofnas (Reference Cofnas2020), for example, contends that convergence on liberal values has occurred not mainly for moral reasons but for reasons of self-interest and economic prosperity that fail to track relevant moral truths.Footnote 5 Further, Hassan (Reference Hassan2019, 151–53) argues that convergence often owes to persuasive speakers propounding false views, members of the public preferring not to hold nonmajority views, and discussants lacking adequate access to nonmoral facts. These are surely important arguments. However, my argument will imply that even if there were little empirically attested convergence of the sort that many moral realists hope for, this would pose far less a challenge to moral realism than many antirealists (and even realists) seem to believe.
1. The actual disagreement challenge to moral realism
By moral disagreements, I will mean conflicting moral views of two or more persons or groups of people. I use the term to refer to the following kinds of disagreement, each of which is a further, dyadic specification of moral disagreement. Moral disagreements can be (a) synchronic or diachronic; (b) explicit or implicit (where agents express or just hold conflicting views); (c) superficial or deep; (d) objective (e.g., in judgment or proposition subscribed to) or subjective (e.g., in one’s affect or attitude); and (e) about general principles or particular judgments.Footnote 6 Moral disagreement thus picks out related but diverse cases of disagreement. By moral agents, moreover, I will mean people who reflect on moral issues and are capable of acting on the basis of that reflection.
Now suppose that people from different cultures fail to agree on many substantive moral questions. Knowledge of such disagreement might incline one to believe that realism does not require convergence. It might attract one to divergentism: the view that if realism is true, rational individuals who are fully informed of the nonmoral facts would not (or need not) have the same basic moral beliefs. And divergentism may be perfectly plausible, as sophisticated theorists have argued.Footnote 7 Alternatively, an awareness of deep and pervasive actual moral disagreement might move one to become an antirealist. This essay is for those who do not find divergentism and nonrealism fully convincing or wonder, in any case, about the plausibility of convergentism.
If realism requires convergence but no convergence is observed, then a rational defender of convergentism must produce a defusing explanation.Footnote 8 Defusing explanations aim to show that disagreement can be resolved by appeal to disagreeing agents’ (a) epistemic defects or (b) lack of information—features that make them nonideal—or their (c) alternative specifications of shared principles, where the presence of (a), (b), or (c) implies that disagreement is not “fundamental.”Footnote 9 The actual disagreement challenge to convergentist realism gets more severe, however, insofar as convergentists also need to explain why disagreement is so widespread and persistent.Footnote 10 Why does disagreement extend to so many people and moral issues? Why does it seem unending and intractable, even between agents with ample opportunity to reflect on and revise their views?
Four prominent kinds of defusing explanation (Doris and Plakias Reference Doris John, Plakias and Sinnott-Armstrong2008, 320) are that moral agents (i) disagree about the relevant nonmoral facts or disagree due to their (ii) own partiality (i.e., bias unjustifiably favoring their interests), (iii) irrationality (e.g., fallacious reasoning), or (iv) different background theories (e.g., different views of personhood).Footnote 11 These categories are complicated of course, and there are subtleties. For instance, as David Brink (Reference Brink1989, 202–3) observes, agents may disagree over relevant nonmoral facts because one or both parties fail to understand the facts, or understand the facts but assess them in different ways but still reasonably. More generally, there are many reasons to think that equally rational agents, and even cultures themselves, sometimes assess fact patterns differently, leading to diverse moral conclusions on specific topics or even general principles.
Challenges to realism, such as Doris and Plakias’s own argument, question the idea, nicely put by Michael Smith (Reference Smith1994, 9), that “[m]oral questions have correct answers … made correct by moral facts.” On realism, if we ask whether X is moral, we can in principle provide a correct answer that adverts to a fact in the world that pertains to X. My argument will target not only Doris and Plakias’s sophisticated antirealist view but convergentist antirealism generally. I will defend realist convergentism according to which moral truths are mind-independent and epistemically accessible to many, as noted in our four-feature characterization of moral realism above.
Doris and Plakias claim to have identified cases of fundamental disagreement that do not admit of defusing explanations. “[F]undamental disagreement,” they observe (Reference Doris John, Plakias and Sinnott-Armstrong2008, 351), “is disagreement that would obtain in ‘ideal conditions’” in which agents make no reasoning errors. Such disagreement threatens realism insofar as “we can be confident that current seemingly intractable disagreements will be dissolved among fully rational agents” (Hasson Reference Hassan2019, 147). But what is an ideal agent? To get one, say Doris and Plakias (Reference Doris John, Plakias and Sinnott-Armstrong2008, 351), “[t]ake any person, and start subtracting reasoning errors, and see what you finish with.”Footnote 12 We end up with a cognitively ideal agent if, I add, our stipulatively perfect reasoner is also fully informed (more on this in sections 2 and 3). However, the authors insist that this should not be taken to imply that empirical data on actual disagreement is irrelevant to the debate over realism. Rather, Doris and Plakias believe, the presence of actual rather than just ideal disagreement is a major challenge to convergentist realism.Footnote 13 Fitzpatrick (Reference Fitzpatrick2014, 165; italics omitted) provides perhaps the best summary of their view in relevant part:
[I]f we can find a significant number of cases of actual moral disagreement that cannot easily be explained away in terms of … defusing explanations (irrationality, partiality, non-moral disagreements, and so forth), then that would cast significant doubt on the realist conjecture that most moral disagreements will be settled under ideal epistemic conditions. Hence, even though the debate may turn on hypothetical rather than actual disagreements, both realists and their anti-realist opponents have a clear empirical imperative: to investigate the extent of actual moral diversity in the world and why it is so persistent. Doris et al.’s project is to show that the convergence conjecture is rendered implausible by the empirical evidence.
To this we can add that turning to the empirical evidence is a very natural move for the antirealist because this evidence might have a special epistemic status as evidence against realism. For one thing, it seems epistemically easier for someone to identify cases of actual disagreement than to speculate accurately about what ideal moral agents would think or do. For another, unlike speculative accounts of ideal agents, anthropologically and social-scientifically informed accounts can employ empirical techniques of independent (third-party) verification. Since Doris-Plakias–style arguments about actual moral disagreement are empirically backed in this way, they have in this respect an important advantage over arguments that purport to identify what an ideal observer might or would think.
Yet it bears emphasis that fundamental disagreement among ideal moral reasoners is perhaps the most troubling kind of disagreement for convergentist realists. Many theorists hold that actual disagreement matters inasmuch as it implies that disagreement would be unsettleable in idealized epistemic conditions. The idea here, put simply, is that if realism requires convergentism, then realism is false if ideal agents would not agree—i.e., “converge.” Now, as many realists believe that much present disagreement “does not (or would not) persist under ideal conditions” (Cofnas Reference Cofnas2020, 3172), Doris and Plakias consider the prospects of realism largely by appealing to claims about whether actual disagreement would disappear in ideal conditions. They write (Reference Doris John, Plakias and Sinnott-Armstrong2008, 319; italics added), for instance:
We can … see how an empirically informed argument about moral disagreement might take shape: Find a case of actual moral disagreement and determine whether any defusing explanations are applicable. If any are, conclude that the disagreement would disappear under ideal circumstances and is not fundamental. If none apply, conclude that a fundamental disagreement has been identified.
Doris and Plakias are not saying that, to identify a fundamental disagreement, we must first understand what maximally rational agents would think about it. Rather, they say, if we can identify defusing explanations for actual disagreement, the disagreement should “disappear under ideal circumstances”—and this, they say, is sufficient to assuage prima facie concern about the threat posed to convergentist realism by disagreement among thoughtful and reasonably unbiased people. The present essay focuses largely on ideal (or more ideal) agents, but in an attempt to identify defusers for actual disagreement.Footnote 14
Doris and Plakias pursue a twofold argumentative strategy. They aim to (1) identify cases of actual moral disagreement and (2) expose as uncompelling certain defusing explanations that imply that these disagreements are not fundamental. I will present Doris and Plakias’s account and then argue that these claimed cases of (actual) fundamental moral disagreement do admit of defusing explanations. If my argument succeeds, it will help clear away a serious objection to convergentist realism.
Doris and Plakias claim to have identified two cases of fundamental disagreement. The first is Honor Culture (HC).Footnote 15 Drawing on empirical work by Nisbett and Cohen (Reference Nisbett and Cohen1996), Doris and Plakias (Reference Doris John, Plakias and Sinnott-Armstrong2008) argue that white members of the American South live in an “honor culture.” This culture upholds certain moral beliefs about the permissible use of violence that people elsewhere in the country strongly tend to reject.Footnote 16 HC concerns, in particular, the kinds of violence that white males in the American South take to be morally acceptable. According to the Nisbett and Cohen study, members of this group are more likely than white males elsewhere in the United States to (i) commit a homicide in response to arguments and (ii) think that, in response to affronts, violence would be “extremely justified,” and nonviolence unmanly. Additionally, (iii) Southern states’ legal systems “allow greater freedom to use violence in defense of freedom and property than do Northern states.”Footnote 17
Doris and Plakias take HC to be a case of moral disagreement between cultural groups in the United States for which there is no defusing explanation. However, Meyers has forcefully argued that HC is not, in fact, a case of fundamental moral disagreement.Footnote 18 I briefly discuss this response to HC before turning to the second, and in my view, more important, case. Some Southerners, the response goes, have irrationally adopted social attitudes that were valuable at times for their ancestors, acquiring and retaining these views largely due to cultural inertia. It was valuable for the “poor, sheep-herding immigrants” populating the South centuries prior to resort to violence or adopt “especially harsh punitive dispositions” (Meyers Reference Meyers2013, 380) when legal enforcement of property rights, itself crucial for inhabitants’ basic welfare, was spotty at best. A willingness to confront dangerous parties with violence using private rather than (sometimes nonexistent) state force—and to signal clearly that one was prepared to do so— served an important protective function (380–81). Fortunately, Southern legal systems are more reliable today, and a disposition to violence in today’s South is scarcely justified on the ground of self-protection.Footnote 19
At issue in HC is specifically whether contemporary Southerners have maintained honor-based dispositions to violence in ways that are epistemically (not just practically) serviceable. If they have, then Doris and Plakias are likely right that there is no defusing explanation, and we have here a case of fundamental disagreement. Meyers does seem, however, to have identified a specifically epistemic flaw insofar as certain contemporary Southerners do not—in a sense soon discussed—weigh the practical costs and benefits of living under the norms of an honor culture as an ideal reasoner would. I will now explain why this is a defusing explanation in HC.
“Even John Locke argued,” says Meyers (Reference Meyers2013, 380), “that in the state of nature, vigilante justice is required and that it would be morally OK to punish theft with execution.” The American South is no longer a largely ungoverned territory in which vigilante justice is beneficial. Yet, attitudes favoring the vigilante justice characteristic of an honor culture persist among Southerners today, evidently as “a matter of habit or cultural inertia” (381). To see why HC admits of a defusing explanation, consider ideal versus nonideal reasoners. In ideal reasoners, there would not be such a lag between an evolving cultural situation and holdover moral principles. Unlike nonideal reasoners, ideal reasoners would update principles quickly (or immediately) to reflect changes in the circumstances of justice (e.g., a shift from social life in an ungoverned territory to social life in the South today). This update would likely be based at least partly on an epistemic principle according to which the costs and benefits of living under a moral principle should be accounted for in determining its truth and applicability.Footnote 20
Now, according to a demanding standard for nonideal agents’ reasoning, the agents in HC are epistemically justified only if they think explicitly about the net value (including the costs and benefits) for their society of upholding the moral norms they uphold. However, nonideal agents need not follow the principle that subject S is epistemically justified in adhering to cultural norm N only if S has shown that N promotes, say, “overall value.” For, in this case, virtually all members of society would not be rational; very few would carefully and explicitly identify and compare all relevant costs and benefits. My view of the requirements of rationality for nonideal agents is more modest. Even if, as nonideal agents, we reason using our System 1 and System 2 thinking (see Sripada and Stich Reference Sripada, Stich, Carruthers, Laurence and Stich2006, discussed below), and even if this reasoning does not approximate that of an ideal agent, we do not lose all justification for our beliefs. After all, we can be epistemically justified despite not being epistemically ideal. In HC specifically, participants could have been epistemically justified in accepting transmitted beliefs on the basis of others’ testimony, but without this justification rising to the level of knowledge that would result from an exacting process of ideal reasoning.Footnote 21 Southerners in HC could have been epistemically justified without explicitly acquiring nontestimonial evidence for their beliefs and evaluating them in the superior ways in which ideal agents would. Many epistemologists today defend precisely such a claim (e.g., Foley Reference Foley1993, 159–60; Plantinga Reference Plantinga1993, 143; Siscoe (Forthcoming); Staffel Reference Staffel2020, 3).Footnote 22
Here, though, lurks a potential trap regarding nonideal reasoning itself—whether the reasoning in HC or reasoning by moral philosophers or any other epistemically nonideal agents. This is the trap of assuming an unduly high bar for nonideal rationality. If we assume a high standard of rationality for nonideal reasoners in HC, this could invite a broad and unpalatable skepticism about the capacity of members of cultures to transmit justified moral beliefs. One strategy for avoiding this trap focuses on the possibility that many moral norms of Southerners in HC are, again, justifiably passed down on the basis of testimony.Footnote 23 A subject can be justified in the testimonial belief that P under two broad conditions. One is in cases where a testifier S1 has justification for P and expresses P to S2, transmitting the justification alongside. A second condition is more controversial but less demanding. An earlier speaker with justification for P can express P to others (or to one other), who can eventually express P to still others, transmitting the justification (see Wright Reference Wright2016, 300–1).Footnote 24 Now cultures typically have some unjustified moral beliefs. One way that members of cultures come to recognize such beliefs is by living with and acting upon these beliefs, and seeing the results (Robson Reference Robson2021). Southerners in HC, for instance, may live in a society where vigilante justice was once valuable. But Southern culture has changed, and Southerners are now living with these beliefs sans many of their former benefits. The benefits have diminished while testimonial transmission of honor-culture ideas persists.
Perhaps it is worth saying a bit more about why nonideal reasoners, such as those in HC, can be justified in their beliefs even when these beliefs differ from those of ideal reasoners. We can be justified on the basis of testimony in thinking that, say, countries exist that we have never visited based on maps or literature or people we have consulted. This is so even if, as nonideal reasoners, we have not ourselves gone through the work of processing the evidence for the relevant geographic claims and eliminated every single error possibility. And, in both this case and HC, our beliefs (and related testimony) can be in error at T2 even if they were accurate and complete at T1. Footnote 25 This makes the aforesaid “trap” less concerning. For much as drivers can usually rely on maps as testimony about highway locations even if maps require updating at times, members of societies can usually rely on testimony about moral norms even if these norms sometimes require revision to account for evolving cultural circumstances. The identification of a gap between nonideal Southerners in HC and ideal reasoners, then, need not lead to a general skepticism. And, more broadly, a commitment to the idea that the cultural transmission of moral norms can fail in some particular cases need not commit one to a sweeping skepticism about the epistemic justification of culturally transmitted moral norms generally. Warranted skepticism about subjects in HC, then, need not invite more general skepticism about all moral agents or all people who reflect on morality, including (hopefully!) us here and now. Even if going through a process of ideal reasoning would show that we are epistemically mistaken in holding to some of our existing moral norms, this does not mean that we are epistemically unjustified in doing so. That would be quite a high bar for justification, one that perhaps only exceptionally strict Bayesians would endorse (see Staffel Reference Staffel2020, 3).
To summarize: ideal agents would follow an epistemic principle according to which, when considering the truth of a culturally transmitted norm or value, one should account for the costs and benefits of living with it. Nonideal agents in HC evidently did not do this. As Meyers suggests (Reference Meyers2013, 380–81), they likely relied instead on the cultural transmission of moral norms that, in this particular case, were no longer epistemically suited to their society’s evolving cultural situation. Further, we can rationally accept this defusing explanation without implying, implausibly, that nonideal agents must always comprehensively think through all of the costs and benefits of each of their culture’s moral norms if the members are rationally to accept them. One might rationally accept a false norm periodically through a process (e.g., testimonial belief transfer) that nonetheless tends to be reliable.
Accordingly, HC seems to admit of a defusing explanation and so is unlikely on balance to be a case of fundamental disagreement. So I shall focus mainly on Doris and Plakias’s second case: the Magistrate and the Mob (MM).Footnote 26 MM deserves special attention because, in my judgment, it not only is a more prominent case in the literature but also constitutes a weightier challenge to realism than HC. My strategy in the rest of this paper focuses on MM but also raises real doubts, I believe, about any attempt to challenge convergentist realism by appeal to actual disagreement.Footnote 27
MM is a hypothetical moral case. “Americans of predominantly European descent and Chinese living in the People’s Republic of China” were asked whether a police chief and judge did something wrong by conspiring to “falsely accuse, convict, and imprison” an innocent man.Footnote 28 Participants considered a scenario that runs as follows:
A member of an ethnic group committed a terrible crime. Enraged, a group of people demanded justice—identification and punishment of the wrongdoer—lest they riot. Unfortunately, the police were unable to identify the perpetrator. The riot would have targeted the perpetrator’s ethnic group, killing some members of the group and destroying property. The only known way to prevent the riot was to accuse and punish an innocent member of the perpetrator’s ethnic group, which the police chief and judge did.Footnote 29
Subjects (or “respondents”) who considered this scenario were asked about moral responsibility. For example, were the potential rioters morally responsible for the decision to convict the innocent man? Did the police chief and judge do something morally wrong by imprisoning the man? Multiple, independent studies concluded that the American and Chinese participants disagreed on whether telishment is morally permissible in MM to prevent an impending riot.Footnote 30 As Doris and Plakias report (Reference Doris John, Plakias and Sinnott-Armstrong2008, 324), “American subjects were significantly more likely [than Chinese subjects] to think that the Police Chief and Judge [sic] were morally wrong to do as they did and significantly more likely to think that they should be punished for doing so.”
MM is thus a good candidate case of fundamental moral disagreement. Doris and Plakias argue that the disagreement in MM is indeed fundamental because there are no known defusing explanations. For it is unlikely that either set of participants was less rational than the other; and, even if one set was, this is unlikely to be germane to MM. Further, participants did not respond in significantly different ways to questions about relevant nonmoral facts, such as whether the riots would end up harming members of the ethnic group. In addition, Doris and Plakias claim (Reference Doris John, Plakias and Sinnott-Armstrong2008, 324–26), partiality is not a viable defusing explanation because “all the subjects are plausibly construed as disinterested third-party observers when filling out their questionnaires” (on which more soon). Doris and Plakias conclude (319) that these key defusing explanations do not apply to MM, so it is false that “the disagreement would disappear under ideal circumstances[.]”Footnote 31 We have here a plausible case of fundamental moral disagreement.
2. Realism defended
Are cases such as MM actual instances of fundamental moral disagreement? Doris and Plakias think so. I find their account powerful but ultimately disagree. At this point in the dialectic, the onus is on me to show why the authors have not adequately considered all potential defusing explanations for disagreement of the kind we see in MM.Footnote 32 I will now develop arguments according to which realism is compatible with MM in particular (in this section) and widespread and pervasive moral disagreement more generally (in the next). These arguments will suggest that ideal—or, at least, more ideal—agents would not think like many respondents in MM did. While there are differing conceptions of ideal moral agency, I shall understand it in a fairly natural way: An ideal agent, let us say, is free from disruptive passions, ways of thinking, and specific thoughts distorted by a bad will; has full information; and is in reflective equilibrium (or, for non-RE theorists, in some other excellent epistemic state). I will also allow, arguendo, that if ideal agents would disagree about X, then there is no fact of the matter about X, so (factualist) moral realism is false.Footnote 33
My view of ideal disagreement is controversial, however. I claim that until we know that an alleged actual disagreement is disagreement to be expected among ideal (or at least epistemically superior) moral agents, we have little reason to think that actual disagreement counts against realism.Footnote 34 One might counterclaim that this gives away the game to the realist because actual disagreement matters as much or more than ideal disagreement. After all, we—real people in actual, historical situations of disagreement—must decide what to do. But if and when we confront others who seem equally epistemically credentialed and well-informed, it would be irrational for us not to either (a) lower our credence in contested moral propositions or (b) give up realism and become either antirealists, if we are realists to start, or relativistic realists, who deny the existence of mind- or culture-independent moral facts.
I am not convinced, however, that when one learns of moral disagreement between people who are equally epistemically well-positioned, one is rationally required to deny that the parties are arguing about a mind-independent fact. First, actual agents are (arguably) rarely epistemic peers (with the same epistemic credentials and evidence for a proposition) and, when we are, we rarely seem to know it.Footnote 35 Second, while the most rational response to epistemic peerhood might be to lower one’s credence in a disputed proposition, there are powerful arguments to the contrary (see, e.g., van Inwagen Reference van Inwagen, Jordan and Howard-Snyder1996; Kelly Reference Kelly, Gendler and Hawthorne2005; Huemer Reference Huemer2011). These arguments typically claim that even when nonideal reasoners know they are peers, they may be rationally justified in not lowering their credence in a disputed proposition but instead “standing pat.”Footnote 36 Whatever one thinks of the rationality of steadfastness as a response to epistemic peer disagreement, a fairly natural response to most actual moral disagreement is to attribute it to the parties’ having different epistemic credentials or evidence and asking what a more (even if not fully) ideal agent would think. Having defended a focus on ideal agents in the context of disagreement between epistemic peers and others, I turn to discussing two common defusing explanations of MM in the rest of this section.
2.a The cognitive cost of moral reasoning: Partiality and epistemic defects in MM
Doris and Plakias claim (Reference Doris John, Plakias and Sinnott-Armstrong2008, 324) that “all the [respondents in MM] are plausibly construed as disinterested third-party observers when filling out their questionnaires.” Relying on this claim, they deny that partiality is a defusing explanation for the disagreement we see. I will argue that partiality is, in fact, a defuser in MM.
To give a flavor for what partiality is, Doris and Plakias (320) cite a statement from Truman Capote (1994), in which the murderer Dick Hickock says: “I believe in hanging, just so long as I’m not the one being hanged.” Hickock’s selective support of hanging humorously displays partiality. The authors go on to observe (320) that “some moral disagreements don’t so much betray deep features of moral discourse as they do the doggedness with which individuals pursue their perceived advantage.” Fair enough. By contrast, practical and theoretical moral thinking that is impartial does not involve the vigorous pursuit of self-interest (including minimizing costs to oneself), but something more like a commitment to other-regard and the earnest pursuit of moral understanding. Doris and Plakias note that partiality is a prominent defusing explanation but think it fails to apply in MM.
I contend, however, that the subjects in MM were not impartial. Rather, on a reasonable understanding they answered the questions partially, in a self-interested way. My argument against the impartiality claim starts with the natural assumption that respondents in MM answered the questions by relying on moral intuitions framed by their culture’s moral norms.Footnote 37 I further suggest that, for reasons of partiality, respondents were unlikely to assess the reasonableness of the culturally specific moral norms on which they relied when answering. These were norms capable of intuitively guiding subjects’ responses if the subjects had internalized them.
Consider, then, what it would have taken for respondents to have been inclined to assess the (moral) norms of their culture that bear on MM. Respondents would not have assessed their culture’s moral norms if assessment required too large an expenditure of cognitive resources or a prohibitively high psychological cost. Insofar as subjects had a sense that the cost of engaging in such thinking would have been prohibitively large, they were partial in their reliance on their culture’s norms. They were disinclined thoroughly to reassess those norms, considering them with the critical distance an ideal epistemic agent would display, due to the cognitive difficulties of such reassessment.Footnote 38 (Note: I use “partiality” here mainly as an epistemic notion, though, in more ordinary usage, “partial” respondents in MM may have had moral shortcomings as well, such as dogmatism or egocentric bias.)
Reassessing norms in a way approximating the thought of an ideal agent would have required participants in MM to ask subtle and perplexing questions, such as: “If I grew up in a very different culture, how might my moral views of MM be different?” And: “Are these views or my current ones more likely true?” Many participants would have a weighty, self-interested reason to eschew such reassessment. Answering questions like these well is demanding and requires considerable cognitive sophistication. It would have been easier for respondents to rely, in epistemically partial fashion, on their own, culturally conditioned moral intuitions—however well the intuitions track any mind-independent moral truths.Footnote 39
We can put this point more generally. Certain large difficulties confront us all as moral reasoners. It is not easy to (i) understand with sophistication the menu of options for moral views within one’s own culture. It is even harder to (ii) articulate a well-reasoned defense of one of these options over the others. It is still harder to (iii) understand the moral views in one’s own country and those in another, quite different country. It is harder yet to (iv) articulate, in addition, a well-reasoned defense of one among this larger set of options. And it is (v) even harder still to understand and articulate one’s own positions among (a) the aforesaid larger set of moral views (those in one’s own country and alternative views in a quite different country) and (b) prominent but different moral views in yet other countries. Now, a sizable numerical minority of persons do incline toward (and some excel at) critical reflection. Nevertheless, it is reasonable to think that, faced with the five challenges above, many subjects in MM (and many of us!) would be epistemically partial. We would opt to defer, somewhat uncritically, to the reigning moral conceptions of our cultures. There seems, then, to be at least as much reason to think that most respondents in MM did not engage in sophisticated critical reflection as there is to believe that they did.Footnote 40 If so, the subjects in MM evidently were partial insofar as they minimized the costs of answering questions by not pursuing considerable reflection on their culture’s background norms and assumptions. A fairly uncritical reliance on their cultural norms impeded respondents’ pursuit of answers rooted in a diversely informed moral understanding.
Ideal agents would not be epistemically partial in this way. As I later argue, ideal moral cognizers in a given culture would not rely on their culture’s norms, or moral intuitions that trace to those norms, without also seriously reflecting on and assessing the truth of these norms. Especially relevant here is that, unlike the subjects in MM, ideal agents can costlessly muster the necessary resources for, say, achieving reflective equilibrium.
But suppose for argument’s sake that many participants in MM did actively seek to reflect critically on their culture’s moral norms. Even so, there is little reason also to suppose that participants were able to do so in a way yielding convergence—even if realism is true. I will defend this claim at length in section 3, but we can now leave behind the previous consideration of partiality to stress that an ability to reflect carefully and thoroughly is perhaps not even very prevalent among actual thinkers (see, e.g., Kahneman Reference Kahneman2011); and, of particular note, many participants in MM were young people with limited reflective experience and relatively few moral and cultural interactions (Machery and Stich, unpublished data).Footnote 41 Participants dozens of years older would have been better able to answer on the basis of additional moral wisdom cultivated by longer, more varied experience. So, we can rationally expect that many of the randomly sampled subjects in MM were unable to reflect well on their culture’s moral norms even if strongly inclined to do so. Contrariwise, ideal moral agents would reflect in a way that is quite clear, comprehensive, systematic, well-informed, and well-reasoned. We accordingly have little reason to expect MM to convey much about the moral judgments of an ideal reasoner.Footnote 42
But what if agents did not rely on their intuitions? While this could spell trouble for my account, Fitzpatrick (Reference Fitzpatrick2014, 179) has persuasively pointed out that respondents in MM unreflectively relied on their moral intuitions, responding to questions on the basis of their moral intuitions, that is, their “fast, automatic, and (often) emotion-laden responses that are the product of entirely unconscious cognitive processes” (169). By contrast, ideal agents would arguably get into reflective equilibrium by bringing into harmony their prereflective, theoretical, and methodological views, and arrive at a rationally stable and coherent set of views.Footnote 43 Participants in MM plausibly relied heavily on System 1, reasoning as discussed in Sripada and Stich’s (Reference Sripada, Stich, Carruthers, Laurence and Stich2006) influential two-system model of cognition.Footnote 44 Participants did not reflect on their moral judgments as an ideal agent would, bringing them into harmony with their moral intuitions, but more likely relied on quick, intuitive judgments internalized from their cultures.Footnote 45 Two-system models of cognition tend to imply that human moral cognition originates in System 1, which enables faster, more automatic and intuitive moral reflections than we get in System 2. And, critically here, System 1 is not widely regarded as sufficient to enable ordinary thinkers to approximate ideal agents. Even Peter Railton’s (Reference Railton2014) notable defense of the sophistication of System 1 thinking, for example, does not imply that such thinking is ideal in the sense of perfect or infallible. Accordingly, insofar as what we can expect from System 1 thinking doesn’t map to the kind of thinking we can expect from an ideal moral agent, participants in MM did not reason as ideal moral agents would.
To say this is not to deny that System 2 thinking can yield problematic reasoning. For instance, moral agents who rely on System 2 can engage in self-serving rationalization. They can be unduly motivated to favor conclusions that prescribe actions partial to their own interests. Further, ideal moral agents might even themselves engage in System 1 reasoning, following their hearts, in a sense, by clinging firmly to intuitive views or reasoning processes. Nevertheless, I suggest that an ideal agent would arguably know the risk that these processes could go awry. The agent would accordingly rely on System 2, or their cognitive counterpart to it, so as to reflectively confirm the conclusions.Footnote 46
2.b An antirealist rejoinder
An antirealist critic might reply that even ideal agents would rely on their cultures’ norms in producing moral judgments; or, at least, their judgments would ultimately trace to those norms, for maybe the information embedded in those norms and how they guide action enables quite competent moral thinking. If so, and if the moral norms vary by culture, perhaps we can rationally expect ideal subjects from different cultures to disagree persistently. Convergentist realism would then be in doubt, as even ideal agents would arrive at different answers to the same moral questions.
My response is to deny that an ideal moral agent would look only to her culture’s moral norms when answering moral questions. The foregoing view proceeds as if such an agent needed an epistemic shortcut for want of time, information, or cognitive resources needed to arrive at a fitting moral judgment of morally permissible behavior for (e.g.) the judge and sheriff in MM. Oddly presupposed here is something like a “bounded rationality” view of ideal moral agents (Grüne-Yanoff Reference Grüne-Yanoff2007). Normal people face serious limits on their reasoning capacity, time, and information; and the theory of bounded rationality as Herbert Simon (Reference Simon1957) intended it was meant to be a satisficing rather than an optimizing or maximizing view—a view for cognitively suboptimal agents like us, not a view of how ideal agents should or do think. We are fallible and limited thinkers, so it is often rational for us to rely on our culture’s moral norms as satisfactory epistemic shortcuts. But this is not rational for an ideal agent. It would be rational for an ideal agent only if the norms on which she relied were rationally justified and the agent knew this by having considered the value of the norms from a perspective external to the norms framework. So, I suggest, the view that an ideal moral agent would look to her culture’s norms and no further (nor elsewhere) when making moral judgments is dubious.
But suppose for argument’s sake that an ideal agent were in a culture possessed of the single, true morality, and that the moral norms in this culture somehow exhaustively expressed the truths of this morality. Even in this anomalously propitious scenario, an ideal moral agent would check her cultural norms for correctness: She would conduct an “external review” of the truth or falsity, and applicability, of the norms. Unlike actual agents, this agent would muster the cognitive resources necessary to look beyond the norms of her culture. She would ask, for instance, “Would my moral views of MM be different if I were raised in a far different culture?” “If they would, what would they be like?” “Is the new response or my current (or another) response more likely true?” We have little reason to think that subjects in MM asked such questions and much reason to think that if subjects in MM relied on their culture’s norms unknowingly, as most people presumably do for survey questions, then they did not think as ideal agents would. Once again, we cannot look to the actions of actual agents in cases like MM to tell us whether ideal agents would converge on a given substantive moral view. Pace antirealists who sympathize with the influential arguments of Doris and Plakias, this puts in doubt MM’s relevance to the assessment of moral convergentism. MM does not indicate fundamental disagreement.
But we can probe further. What if the subjects in MM somehow engaged in the same kind, quality, and amount of thinking that an ideal agent would? Even on this likely false assumption, the Doris-Plakias–style antirealist would be saddled with a troubling worry. I call it the generalization worry: If subjects did not rely on their cultures’ norms, then their responses are both hard to explain and unlikely to generalize. To see why, suppose you were a respondent in MM. And suppose that rather than having you quickly report your moral view on X, researchers encouraged you to think long and hard about X, to the point that you eventually reflected so much that you arrived at a view of considerable sophistication. Now notice: having engaged in such reflection, you are more likely to have employed premises not in the common stock of those widely held in your culture. Your thinking will thus likely diverge from your culture’s thinking more broadly, if not in your conclusion then at least in the complexity of your reasoning, becoming more independent from it. Your response to MM will then be more like an anomalous or inexplicable data point than one known to be representative of what members of your culture tend to think (if a tendency obtains).
This section has two main results. (1) Assessing the moral views and norms prevalent in one’s society “if that’s where the argument goes” requires substantial cognitive effort. Thinking in this way is something we can expect of ideal agents. It is not something we can expect of a random set of surveyed persons. (2) Even if respondents in MM did critically reflect on their culture’s moral norms, this would not count as weighty evidence against convergentist realism. Anyone in MM who reflected like an ideal agent would have engaged in independent thinking that was complex, possibly anomalous, and hard or impossible for empirical studies to track.Footnote 47
3. Three reasons why realists should expect deep and persistent moral disagreement
This final section develops three more defusing explanations. These provide further reason to resist any antirealist implications of actual moral disagreement in MM and, indeed, more generally. The defusers that I propose concern how individual moral views develop over time, group morality develops within a community, and persons model the moral world.
3.a Disagreement and Time 1: Individual development of moral views
MM was a time-slice experiment: the idea was to ask people hypothetical questions in order to assess their moral views at a point in time. Experimenters could then confirm or challenge the results by repeating the experiments, say, every year. Rather importantly, though, no such study has been done over several decades comparing longer term moral patterns. Yet, despite this absence of empirical work, theoretical responses to MM seem to assume not only that the experiments can capture patterns of moral development at a time slice, but also that long-term moral development will conform to such a pattern in ways incompatible with realism. This assumes a controversial view of moral development across time. In this subsection and the next, I argue that this assumption is unfounded.
The first of the three defusing explanations concerns the intrapersonal development of moral views (the next subsection discusses interpersonal development). We can extend Fred D’Agostino’s (Reference D’Agostino2010) work on scientific communities of inquiry to show why popular views of morality might diverge across societies at a given time, even if both moral realism is true and convergence will eventually obtain. D’Agostino illuminatingly shows how a “wave” can sweep through a community of individual scientific inquirers. The idea is simple but powerful: a theorist adopts a view and then develops it theoretically such that it becomes attractive to another theorist. This second theorist then develops the view further, making it attractive to a third, previously skeptical theorist, who adopts it as well. This process continues indefinitely with the core view remaining intact but being refined over time to better fit inquirers’ particular desiderata for acceptance.
The development of moral views, I suggest, often conforms to a similar pattern. Individuals develop their moral views within a community as follows: Person A holds a moral view and improves it by (e.g.) rendering it more internally consistent. Person B, noticing the improvement, adopts the view.Footnote 48 A and B then discuss the view and refine it further. Eventually, the view becomes attractive enough for C to adopt. The process continues indefinitely until person N, say, is the last to adopt the view.Footnote 49 All along, the core view remains intact but is continually revised to fit moral agents’ own desiderata for acceptance.
Now, this process takes time, and at any given point (time slice) two persons—A and C, say, or A and N—might hold very different views. Their disagreement may or may not be resolvable at that time, as many theoretical disagreements often are not for scientific inquirers at a particular time slice.Footnote 50 But this does not mean the disagreement will never be resolved. For if moral development fits the pattern that D’Agostino describes of theoretical development in the natural sciences, we can expect groups of moral inquirers to hold diverse views at different times, with the views enjoying mixed levels of popularity or uptake in diverse groups at alternative times. This seems entirely possible in MM and other cases of time-slice disagreement—even if realism is true.
This pattern seems to have obtained in various social-moral situations historically. For example, some countries abolished slavery far earlier than others. Before the practice was widely abolished, there were deep and lasting moral disagreements about the moral permissibility of slavery, both within and between moral communities. These differences eventually vanished. And differences in views on, say, the moral permissibility of racism and sexism have also largely vanished; when voiced in the mainstream, these views tend to be met with opprobrium by diverse respondents. There are still many racists and sexists in the world today, alas, and even slavery of various forms remains a formidable global problem. But in virtually all extant cultures, vocal supporters of these egregious practices are now outliers. Real moral victories have been won. And real moral convergence, often long thought utterly unachievable, has occurred (see Huemer Reference Huemer2016, on which more later).Footnote 51 The upshot here is that individual moral disagreement at a time-slice is indeed compatible with convergentist realism.
3.b Disagreement and Time 2: Group development of moral views
We now turn from the moral development of individuals within communities to patterns of group moral development. Discussing moral convergence, Smith (Reference Smith1994, 187) observes that theory construction takes “reflection and conversation,” and “it might also take time and effort to convince anyone else” of the truth of one’s theory. Even if you had the best moral theory and knew you did, it would likely take you much time and effort to convince others of its merits. The requirements of time and effort deeply affect how groups learn about, evaluate, and settle on their moral views, often one member at a time. What would it be like to apply these insights to the case of moral disagreement between American and Chinese subjects in MM?
The argument of this subsection proceeds in three steps. First, grant for the sake of argument that a version of realism is true. Second, consider a series of five substantive moral views, V 1 … V 5, that agents hold at (e.g.) six different times, T 1 … T 6. On the schema below, two cultural groups start to think about morality at the same time, T 1. The groups could respectively hold the five stipulated moral views in the following temporal sequences (note the changes from T 2 to T 6):
Third, recall that sifting through information and learning about morality takes time. Even if the two groups follow exactly the same sequence of moral development from T 1 to T 6 just at different times, they could end up holding far different views in a large majority of time-slices: T 2 to T 5. One group might even lag just a single step behind the other for most of the time, as above. And even so, the groups could end up with the same view at T 6, perhaps even holding it stably thereafter. My question: At which time-slice, so to speak, were the Chinese and American respondents in MM? Answer: We don’t know.Footnote 52 Does this lack of knowledge matter? Yes. Quite so. It matters if we want to see, as we do, whether the groups will permanently diverge or eventually converge.
From the perspective of an empirical observer, V 1 might differ considerably from V 2, V 2 from V3, and so on. But even if V 5 were the true moral view and the groups were in fact converging on it, empirical observers of disagreements in MM or other venues would generally be oblivious to this fact. Realism would be true while disagreement consistently appeared deep and intractable. Culturally different groups on the same sequential moral paths would long appear not to be. It is very plausible, however, that if different cultures will eventually converge upon moral truths, the process will involve diverse developmental trajectories and take considerable time (Smith Reference Smith1994, 187). So we should expect short-term disagreement whether or not the agents are converging over the long term. If so, actual disagreement at a given time would not be clear evidence against realism.
It is worth distinguishing the two preceding defusing explanations from the earlier one regarding epistemic partiality. Unlike the defusing explanation from partiality, the defusing explanations from individual and group moral development do not imply any sort of immediate epistemic failure on the part of the agents involved. The agents are not overly biased or self-interested, for instance. Nor, by stipulation, are they consciously or subconsciously unwilling or unable to reflect on their culture’s moral norms. Empirical studies can readily miss the fact that the selfsame conscientious moral reasoners may reason in different directions at different times. This matters greatly for cases like MM that attempt to tease out cross-temporal moral lessons from observations made at a time-slice rather than, say, over decades in which individuals and groups develop, refine, supplement, and, at times, supplant their moral views.Footnote 53
3.c Moral modeling in the development of moral views
A comparison between moral agents and scientific modelers further suggests that actual moral agents act quite differently from epistemically ideal agents. This implies that we cannot infer safely from the behavior of subjects in MM that ideal agents would disagree as participants in MM did. The blockage of this inference blunts the force of critiques of convergentist realism based on moral disagreement found in experimental settings (e.g., MM) and perhaps most actual social settings. Let me explain.
Philosophers of science John Matthewson and Michael Weisberg observe (Reference Matthewson and Weisberg2009, 169) that “[d]espite their best efforts, scientists may be unable to construct models that simultaneously exemplify every theoretical virtue.” When constructing theoretical explanations of the world, scientists face tradeoffs such as the tradeoff between precision and generality. More precise scientific “models will tend to apply to fewer cases when one evaluates them with higher standards of fidelity or intends that they capture more aspects of the target phenomena” (186). Ceteris paribus, a model that characterizes a phenomenon in less detail will be less precise. A more general model will explain more cases or apply to more phenomena, but often at the cost of forgone precision. By contrast, a model that is quite precise but not very general may cover few cases, say Matthewson and Weisberg (186), but yield theoretical results “that mimic the target very closely.” Accordingly, the precision and generality of a scientific model will often depend on the goals of the modelers.
Might actual and ideal moral agents model the moral world much as scientists model the physical world? And if so, might moral agents and scientists face similar model-theoretic tradeoffs? Thomas Kuhn famously observed (see D’Agostino Reference D’Agostino2010, 86) that scientific results depend markedly on how much weight theoretical scientists give to values such as accuracy, plausibility, and consistency. Yet, he also stressed, scientists who embrace these values and rely on the same method will not necessarily achieve the same results for diverse scientists assess and weigh the relative merits of even shared values (e.g., precision, generality) differently.Footnote 54 The same lesson, I will now argue, holds true in the moral domain, including for cases such as MM.
Like scientists, actual moral agents are epistemically nonideal, and this often necessitates a reliance on moral models. Diverse agents employ different theoretical desiderata in developing their own moral views. They care to differing extents about the coherence of their views, how and how far their views ought to be empirically supported, how well their views harmonize with their pretheoretical intuitions, and much else. Like scientific modelers, moral modelers are far from omniscient in the relevant domain; they too must do the best they can given their limited theoretical capacities. This is a weighty reason to expect actual agents’ moral views to deviate from ideal agents’ views. In MM, for example, an ideal agent (to whose thinking defusing explanations do not apply) would know which values were at issue (e.g., protecting the innocent, preserving ethnic rights, preventing a harmful riot), which nonmoral facts were relevant (e.g., the likely social effects of framing different people), and how to tradeoff between values if and when trading off is required.Footnote 55 So understood, an ideal agent in MM would either need no model of morality or would have a better moral model than either group of respondents had in MM. We cannot conclude, then, that ideal moral agents would disagree about MM because actual agents did.
A final comparison between moral agents and scientific inquirers merits note. While natural scientists disagree on many aspects of the activity, development, and nature of the physical universe, they strongly agree that, for instance, electromagnetism, gravity, and the strong and weak nuclear forces are fundamental to it. Similarly, notice in MM that we do not have maximally deep moral disagreement; rather, in a notable sense, basic moral understandings of human association remain intact. For presumably no one thinks, say: “Yes, let’s frame and punish innocent human beings for fun!” The disagreeing parties would equally reject this proposal. Accordingly, much apparently “fundamental” moral disagreement in MM and generally thus seems to concern instead how best to instantiate or apply the same basic principles of morality. Some argue (e.g., Wall Reference Wall2016; Robson Reference Robson2020), for instance, that basic principles of political morality are diversely realized across different political traditions. At issue here is typically not whether principles (or values) such as political freedom and equality apply, but how they do, especially where trading off between basic moral principles is required in practice. I suggest, then, that even if ideal agents would disagree about MM, this disagreement would presumably likewise be more about how best to apply (fundamental) principles already agreed to apply.
For suppose, hypothetically, that ideal agents disagreed about which course of action in MM—imprisoning an innocent person or allowing a deadly riot—is more moral (assuming the options are not equally moral). These ideal agents would, for all we do know, mutually accept the principle that it is quite morally important to protect innocent lives. Where they would disagree would be on the shallower moral issue of how to balance (a) protecting potential innocent victims of a riot from harm or death and (b) protecting a potential innocent person from unjust imprisonment. This disagreement is at least to some extent about how best to apply a basic, uncontested principle of the rights of the innocent; it is not necessarily disagreement over the truth of that principle itself. If ideal agents disagreed about MM or were indeterminate about how best to proceed, this would not necessarily imply the falsity of moral realism. The agents might yet agree, more importantly, on more foundational moral principles.
3.d The ”mere possibilities” objection
One might worry, however, that the foregoing view about moral views developing over time trades too much in possibility. Yet I have argued for a fact: the noteworthy fact that robust time-slice disagreement does not alone pose a challenge to realism. And to argue that time-slice disagreement need not be a sign of long-term divergence in moral views is to make progress on a thorny philosophical issue. Here the following fact matters greatly: Metadisputes about morality concern large and diverse human societies that consist of quite complexly interrelated persons with (of course) human brains—that is, the most complex objects in the known universe.Footnote 56 To make claims about patterns of moral thought and speech that do or might obtain among millions of persons, each with brains of 85–100 billion complexly interrelated neurons, is thus to enter upon social-psychological terrain of virtually unfathomable complexity, with societies themselves quite possibly being the most complexFootnote 57 multiobject entities in our massive universe, itself virtually unfathomably complex.Footnote 58 While Doris, Machery, Plakias, Stich, and others have admirably compiled and examined important data on actual disagreement, we all (antirealists and realists alike) are left, in the face of great social complexity, to speculate to some extent on the normative import of the data. My argumentation establishes, at the very least, that we do not know whether people in different groups are on a converging long-term path, and that they could well be. And to know what we do not know is to gain a valuable sort of knowledge.
Still, suppose that despite my argumentation, antirealists persist in having, as Vavova (Reference Vavova2014, 315) puts it, “good independent reason” to think realism is false.Footnote 59 The potentially converging trajectory of moral development just sketched might then be reasonably rejected if it is at odds with those independent reasons. Even this is not clear, however. For what is being intuited here is a justification for lower confidence in a potential convergent moral trajectory on the basis of one’s antecedent (independent) reasons to uphold antirealism. But this seems no more justified than intuiting a justification for lower confidence in those antecedent (antirealist) reasons themselves, given a plausible account of how actual moral disagreement is compatible with realism. I have just provided such an account, one of how cultures could converge on moral truths via differential moral trajectories, yielding disagreement throughout much—indeed, most—of the process. As the saying goes, one person’s modus ponens is another’s modus tollens. The potentially surprising claim that, even on moral realism, deep and widespread disagreement is rationally expectable puts those who claim that divergence is evidence against realism under pressure to show why, for instance, it could not be part of a historical trend that is shifting or will shift toward convergence.
And one recent shift towards convergence has been particularly striking.Footnote 60 Huemer (Reference Huemer2016, 1990–93) argues that recent centuries have witnessed a robust convergence on the immorality of broadly “liberal” values against, for instance, torture (1990), slavery (1990–92), racism and sexism (1993), colonialism (1993), and much else besides (see also Smith Reference Smith1994, 188). Now it would be Whiggish, of course, to assume that humanity inevitably moves towards better moral views on every issue or even most. However, Huemer’s argument puts antirealists under further pressure to show that convergence has not recently occurred on various, previously quite controversial issues. While scholars (Hassan Reference Hassan2019, Cofnas Reference Cofnas2020) have recently developed important challenges to Huemer, it should never be forgotten that in (say) 1700, the complete abolition of chattel slavery seemed shockingly unlikely—and, sad to say, even immoral. The immorality of slavery is now taken for granted virtually everywhere, and the very idea of debating slavery’s moral status would now be rightly regarded as morally monstrous.
4. Conclusion: countering the actual disagreement challenge to convergentist realism
A prominent antirealist argument claims that (1) realism appears to require convergence, but (2) there are many documented cases of divergence. By modus tollens, the line goes, realism seems false. I have exposed vulnerabilities in this basic line of argument. If realism requires convergence among ideal agents, this does not imply that we ought, rationally, to expect the views of actual agents to converge at any given time.Footnote 61 I have argued, on the contrary, that we should expect robust time-slice divergence whether or not there is eventual convergence. This is largely because actual moral agents, including subjects in MM, are (i) epistemically partial in a way that unreflectively favors their culture’s moral norms, (ii) otherwise cognitively deficient, (iii) apt to diverge from each other, at a given time, in terms of individual and group moral development, and (iv) apt to build different moral models. The upshot of my account is that empirically attested, time-slice disagreement in MM and more generally does not unsettle realism. If the threat to realism from actual moral disagreement is now considerably deflated, this puts the ball back in the court of antirealists who focus on disagreements among ideal moral agents rather than on disagreements among agents like us, with our deep and ineluctable cognitive limitations.
Acknowledgments
I thank Michael Bukoski, Karina Robson, Connie Rosati, Wes Siscoe, and three anonymous referees for their generous feedback on earlier versions of this paper. I also thank Edouard Macherie and Stephen Stich for kindly sharing their unpublished data with me.
Gregory Robson is an assistant professor of philosophy at Iowa State University. His main research areas are ethics (including business and technology ethics) and political philosophy.
