R2.1. Irrationality

For now, let us set aside particular kinds of bad influences (more on that later) and consider more general worries about irrationality in ethics. Landy admits, at least for the sake of argument, that “better reasoners are more likely to arrive at well-founded moral beliefs than are worse reasoners.” But he suggests that many of us are bad reasoners in general and thus lack well-founded moral beliefs. It is tempting to dismiss this as a radical form of skepticism about our general reasoning capacities, which I set aside as outside the scope of the book (see especially p. 22). But Landy astutely pitches his argument as one “for a moderate amount of pessimism,” which “cannot be dismissed as merely radical skepticism.”

My response is two-pronged. First, since pessimism comes in degrees and my optimism is only of the cautious “glass half full” sort, I am happy to concede a good deal of pessimism. However, second, let me say a few things to suggest that the pessimism about reasoning generally only goes so far.

Landy’s pessimism is based primarily on evidence of our general poor performance on cognitive tests, particularly the Cognitive Reflection Test (CRT). There are multiple versions of the CRT now, but the one used most often by far involves what (in the United States at least) are called word problems in grade school math class. For example, one of the three questions on the original CRT is “If it takes 5 machines 5 minutes to make 5 widgets, how long would it take 100 machines to make 100 widgets?” (Frederick Reference Frederick2005). The mix of numbers, repetition, and counter-intuitive tricks in such word problems makes most people's eyes glaze over, and consequently they do not exert the cognitive control to carry out the right inference. The struggle with word problems, however, does not necessarily reflect irrationality generally or a general unwillingness to reason.

Consider another version of the CRT mentioned in the book (p. 71), which downplays the numbers and deals with more ordinary scenarios. For example, one question reads: “Emily's father has three daughters. The first two are named April and May. What is the third daughter's name?” (Thomson & Oppenheimer Reference Thomson and Oppenheimer2016, p. 101). The intuitive answer is “June,” but the right answer after a little reflection is “Emily.” Online participants correctly answered on average a little more than two of the four questions, which is only slightly higher than the original CRT. Nevertheless, it does suggest that people reason better with more ordinary problems. And yet the data that Landy draws on regarding performance on the CRT rely almost exclusively on the original CRT with the mind-numbing number problems.

Another reason for optimism lies in a comparison with research on the famous Wason selection task. This problem, which originally involved two-sided cards, requires one to determine when a conditional statement is false, such as “If a card has an even number on its face, then it is red on the other side.” Which card(s) should you turn over to determine whether the conditional is false? Few people are successful at this seemingly simple task. As it turns out, however, most people can solve it easily if the context is one we are more equipped to deal with, particularly when it involves conditional social rules, such as “If a man eats cassava root, then he must have a tattoo on his face” (Cosmides & Tooby Reference Cosmides, Tooby, Barkow, Cosmides and Tooby1992). Even children as young as three can navigate a Wason-style selection task if framed in terms of a norm, such as “If a mouse is squeaky, it must stay in the house” (Cummins Reference Cummins1996). So, even if we are no good at tackling certain mathematical word problems, we may be much better when reasoning about their social or normative counterparts.

Finally, the CRT tests for reflective inference, in which one must slow down and consciously apply the steps in a process of reasoning. However, this only represents one kind of reasoning. As I emphasize in the book (especially Ch. 3), much reasoning is unconscious, automatic, and intuitive. Perhaps Greene (Reference Greene2014) is right that the resolution of controversial moral problems requires engaging reflective “System 2” reasoning. But even Greene acknowledges that in everyday life we should by and large rely on intuitive moral reasoning. And we do have evidence that intuitive reasoning is often both sophisticated and reliable (see, e.g., Gigerenzer Reference Gigerenzer2008; Nichols et al. Reference Nichols, Kumar, Lopez, Ayars and Chan2016). Indeed, in ethics, moral judgments associated with more reflective reasoning are correlated with more anti-social and egoistic moral views (Kahane et al. Reference Kahane, Everett, Earp, Farias and Savulescu2015). So I caution against using (only) reflective reasoning as a test of good reasoning abilities, especially given that being adept at reflection and consciously articulating a justification often just helps one to better rationalize a bankrupt position (see, e.g., Kahan et al. Reference Kahan, Peters, Wittlin, Slovic, Ouellette, Braman and Mandel2012; Kunda Reference Kunda1990).

Indeed, reflective reasoning is sometimes good, sometimes bad. Often we are unable to accurately identify and articulate our reasons, which can lead to post hoc rationalization. Clark & Winegard are not themselves pessimists, but suggest that without the right emotional intuitions post hoc rationalization can make moral debate full of self-deception. However, this is not always the case. Initially inaccurate attempts to explain one's own choices can ultimately lead to insights and improve one's behavior (Summers Reference Summers2017). For example, I might not fully understand why I stop eating meat, but that forces me to research factory farming and to be consistent in whatever analysis I come up with (compare similar points made by Holyoak & Powell).

In general, it is important to keep in mind that an inability to identify and articulate one's reasons does not imply a lack of reasoning or poor reasoning. Children seem to reason in sophisticated ways well before they are able to consciously identify and articulate their reasons (see, e.g., Gopnik Reference Gopnik2012). Once when my daughter was 5 years old she asked, “Dad, did you pack my toothpaste [flavored] jelly beans?” Trying to be dead-serious, I said “Well, no … I ate them all. Sorry.” She replied: “No, you didn't! You don't like them.” The fact that she added an explanation was noteworthy. When she was younger, she would have called my bluff by just saying “No you didn't!” without offering a reason. The ability to articulate the reason came later.

In sum, poor performance on tricky math problems does not necessarily warrant pessimism about reasoning in general. Moreover, we have positive evidence that we are often even better at logical reasoning in moral contexts involving rules. Finally, reflective reasoning is only one form of inference, and it is not generally unreliable or reliable, for it can lead to both good and bad reasoning. Still, I take many of Landy’s points which contribute to my optimism only going so far. In particular, it is true that laziness can prevent many of us from engaging in effortful reasoning which is important in the moral domain, where one must exert great effort to carefully and charitably consider the opposition.

R2.2. Rationalizing our biases

Even if our general reasoning capacities are fairly sound, one might worry about our starting points. In adolescence or adulthood, our independent reasoning capacities may be up against powerful biases we have absorbed from our cultures and institutions, from misogyny to materialism. Reason can promote consistency in ethics (Ch. 5; Kumar & May Reference Kumar, May, Suikkanen and Kauppinen2019), a point well-illustrated by Holyoak & Powell, and I appreciate their proposal that coherence-based reasoning helps explain the difficulties with debunking most moral cognition or motivation. But, a skeptic may retort: “Put garbage in and you will get garbage out.” We have to consider the materials that mature moral reasoners begin with.

This is the sort of message one might take away from several of the commentaries. Rottman, for example, recognizes all of the evidence that children are like little scientists, unconsciously reasoning and experimenting to learn about the causal and social structure of the environment. However, he points out that as children we are, despite some bouts of altruism, “particularly egocentric” creatures who start with some psychological dispositions that are adaptive even if not moral, such as tribalism and prejudice. Moreover, children are primed to uncritically learn from their cultures and as a result can, as Rottman puts it, form “novel moral beliefs in irrational and undiscerning ways.”

Rottman and I may not disagree greatly. He acknowledges that children often engage in unconscious reasoning that is quite sophisticated. Rottman only defends a “middle ground” that maintains a “healthy dose of pessimism,” something I welcome, as well, especially because it can foster moral progress and combat complacency. Because children are not yet mature moral agents, I focused on adulthood. Still, I take the point that our developmental origins shape adult moral reasoning (see also Carpendale & Wallbridge; Narvaez), and often not for the better.

We may find evidence of this in Zheng’s commentary. Her concern is that social cognition generally is so riddled with bias that most people are likely to use reason to justify morally reprehensible attitudes and actions, such as racism, sexism, and xenophobia, especially given our unjust social conditions. Many people come into adulthood with homophobic tendencies, for example, that were likely inculcated by their culture, society, and perhaps evolutionary history. For many people, forming enlightened attitudes about homosexuality requires great effort and many do not succeed. Similarly, there are unfortunately powerful amounts of racism, xenophobia, and dogmatism that good reasoning has to work against. Moreover, Zheng rightly worries that the pervasiveness of motivated reasoning means we are likely to just rationalize many of the morally reprehensible attitudes we start with.

Like other calls for some pessimism, I concede it as far as it goes; and, again, it certainly goes farther than the tone of the book suggests. But it is important to remember that my optimism is only about the role of reason in moral psychology and our basic modes of moral thought and motivation. Of course, as Zheng makes clear, reason can justify terribly inaccurate ideologies, even ones that are woefully unjustified. However, whether our concern is moral judgment or social cognition generally, I do argue that many illicit forces, including implicit biases, are not as powerful as they are often thought to be (Chs. 4 and 9). Of course, as I say, little biases can add up to a powerful effect on society as a whole, just not so much on individual beliefs and desires. So I agree with Zheng that a badly ordered society impairs reasoning and acting well, but then moral reasoning is not rotten to the core but rather situated within unjust societies.

Second, I suspect that correcting structural injustices will require improvements in individual rational capacities, not mere feelings (conceived as distinct from reason). To combat misogyny, for example, we cannot just get men to feel more positive toward women. As Kate Manne (Reference Manne2017) has recently pointed out, sexists can still care sincerely for their mothers, sisters, friends, and coworkers, all while perpetuating patriarchal attitudes and social orders at home and beyond. A loving husband can still expect his spouse, daughters, and women politicians to be “attentive, loving subordinates” (Manne Reference Manne2017, p. 49) and mete out punishment if they fail to live up to those expectations. The problem here is not mere feeling or affection but unconscious patriarchal ideologies and social norms. Now, such problems may have less to do with individual moral cognition than entrenched cultural norms and institutions. But individuals may bear some responsibility for the structural injustices, which one cannot combat without reasoning that the ideologies and structures are flawed (Madva Reference Madva2016; Zheng Reference Zheng2018).

In sum, as a cautious optimist, I am a good deal pessimistic. The more one studies the science, the clearer it is that we need to be humbler in our controversial moral views and recognize our ability to rationalize our egocentrism and prejudice. I agree with Zheng that these are the most troubling and pressing sources of moral irrationality, but as she also emphasizes this is well known from history. As far as the psychological science, I remain doubtful that we can ground much pessimism in incidental disgust, evolutionary pressures, framing effects, automatic emotional heuristics, and the like. Perhaps now then is a good time to defend my optimism about those potentially bad influences that have received so much attention in moral psychology.

R2.3. Bad influences

My optimism about moral cognition and motivation is due primarily to the failure of wide-ranging attempts to debunk them. I framed the skeptical challenges in the simplest of terms to highlight their key empirical and normative premises:

1. 1. Empirical premise: Moral cognition/motivation is mainly based on some factor.

2. 2. Normative premise: That factor is morally irrelevant, extraneous, etc.

3. 3. Normative conclusion: Moral cognition/motivation is improperly influenced (and so unwarranted, improper, etc.).

One drawback of this oversimplification is that it obscures an important issue, which Demaree-Cotton nicely draws out. It is not necessarily problematic that a belief or motivation has a bad influence, even if it is substantial. A pleasing smell or energetic background music may be irrelevant to my task of determining the appropriate punishment for a defendant, but those same factors could help focus my attention on morally relevant facts of the case. As I point out in the book (e.g., pp. 13, 28, 31, 48, 71), emotions in particular can help draw one's attention toward (or away) from relevant information. Demaree-Cotton is quite right to note, however, that my frequent talk of “substantial influence” can suggest (wrongly) that debunkers need only point to an extraneous cause whether or not it also ignites a reliable process of forming one's moral belief or motivation. The better phrase – which I do employ in the official statements of the Debunker's and Defeater's Dilemmas – is that the relevant judgment or motive is “mainly based” on the bad influence. That phrase at least suggests more clearly that the resulting judgment or motive is based on the bad influence and not also a good influence.

As Demaree-Cotton notes, if we emphasize this point, then it is even harder to draw skeptical conclusions from the data. Even if, for example, incidental disgust substantially influenced a wide range of moral beliefs, debunkers would still have to show that this emotion does not draw one's attention to morally relevant information. Still, it seems reasonable to assume that many if not all bad influences – incidental disgust, genuine framing effects, implicit biases, and so on – do not generally beget good influences. I at least tended to grant my opponents this assumption for the sake of argument.

In the book, whenever there was a concern about nearby good influences, I typically addressed the issue in relation to the normative premise of the debunking argument, which essentially concerns whether the influence is actually bad or good. Consider, for example, my discussion of framing effects in chapter 4 (more on that soon). The order in which information is presented can seem like a bad influence on moral belief, but it is perfectly appropriate if it leads to rationally updating on new evidence (see Horne & Livengood Reference Horne and Livengood2017). Rather than saying this is a bad influence that generates a good one, I typically treated the seemingly problematic influence as not so bad after all.

Speaking of problematic framing effects, McDonald et al. take issue with my claim that meta-analyses suggest that such effects on moral belief are generally small. Part of their worry is that I rely primarily on only one meta-analysis by Demaree-Cotton (Reference Demaree-Cotton2016), which McDonald et al. believe is flawed. Although they do acknowledge that I also rely on another meta-analysis (Kühberger Reference Kühberger1998), they object that it concerns framing effects “on risky choices rather than morality.” But many of the risky choices examined in that meta-analysis are morally relevant ones, such as tax evasion, contract negotiations, game-theoretic social dilemmas (e.g., public goods games), and public health policies such as the famous “Asian disease problem” (from Tversky & Kahneman Reference Tversky and Kahneman1981). Moreover, my reliance on Kühberger's meta-analysis is important because it suggests that one of the most substantial framing effects (the disease study) is actually an outlier, not representative of similar decisions.

McDonald et al. also object that Demaree-Cotton’s meta-analysis is limited and that they are currently preparing a larger analysis of framing effects on moral judgment. That is very welcome news, and I look forward to seeing the final results. Until then, we must work with the data available. What can we conclude so far?

McDonald et al. urge us to recognize that framing effects on moral judgment do not just arise for order of presentation and the like. They point to studies showing that moral judgments are influenced by videos, sleep deprivation, and social milieu. But these are often morally relevant factors. It is important to realize, as we saw previously, that a factor can seem morally relevant in the abstract but is not problematic when understood in context. Consider sleep deprivation and compare its effects on mathematical judgment. Suppose, as is plausible, that studies show that sleep deprivation dramatically affects one's solutions to math problems. There is no doubt that sleep deprivation tends to negatively affect mathematical judgment, but the question is whether this general factor of amount of sleep is a defective way of forming mathematical beliefs. Again, it may seem so in the abstract but amount of sleep – in particular, sufficient sleep – does plausibly ignite mathematically relevant cognitive processes. Indeed, it is a truism that quality and quantity of sleep directly impacts general cognitive functions, including those relevant to quantitative reasoning. Similarly, when properly described, many factors affect one's perception of morally relevant considerations, including videos, amount of sleep, and social milieu.

Another issue raised by McDonald et al. concerns Demaree-Cotton’s finding that roughly 80% of moral judgments remain unchanged even when subjected to framing effects. The commentators object that 20% is large enough for skepticism to creep in. Even if the effects are small, they can add up for any single individual and ultimately sway some of their moral judgments for arbitrary reasons. The exact structure of this idea could take different forms.

On one reading, this is a version of what, in chapter 5, I call the “generalization worry” (p. 112), which can be found in the work of others, as well (see, e.g., Doris Reference Doris2015; Rini Reference Rini2016). I address the generalization worry in chapter 5 but also in chapter 9, where I discuss this move among those who attempt to debunk or “defeat” moral motivation. In both places, my response to the generalization worry is essentially that it fails to demonstrate evidence of unreliability (I take it that 80%, e.g., is reliable), which is necessary for a debunking argument.

One could retreat to merely raising the hypothetical possibility of unreliability. McDonald et al. do worry that “we as individuals often do not know whether we are in the unreliable 20% or the reliable 80%.” But such “skeptical hypothesis arguments” are notoriously weak (May Reference May2013b). If swallowed, they lead to radical skepticism well beyond the moral domain. Some brilliant philosophers have defended such forms of argument and some have embraced the radical skepticism that follows. But few of my opponents aim to go that route. Indeed, if they did, the empirical evidence would be entirely irrelevant, for as Descartes showed us long ago one can spin out skeptical hypothesis arguments day and night while meditating in the armchair. At any rate, my concern is whether moral cognition is unreliable, not whether unreliability is merely possible.

McDonald et al. might be offering instead an argument from disagreement. They write that “one cannot know that one is a reliable moral judge, whereas others are not, unless one has some reason to believe that one is special in some relevant way.” I entirely agree, provided we do restrict the skeptical conclusion to whether “moral judgments in controversial cases can be justified” (my emphasis), as I do in chapter 5 when embracing a selective debunking argument from peer disagreement (more on that in the next section).

It is quite important to distinguish the different forms of argument here, which point to either: the mere possibility of error, likelihood of error, or the prospect of error from peer disagreement. An analogy may help. Consider my various beliefs about the capitals of countries, such as the belief that Lima is the capital of Peru. Of course, it is merely possible that I am wrong, but that is no reason to suspend judgment. Now imagine I learn that 20% of my geographical beliefs are formed in a faulty way, such that if those 20% are accurate it is a kind of fluke. Should I suspend judgment about Peru's capital? I do not think so. Suspension would be required if I learned that 80% of my geographical beliefs are suspect, but not 20%. Matters change greatly, however, if we introduce epistemic peers. Suppose instead that I learn that 20% of my peers disagree with that particular belief about Peru. If these are well-informed people who I have no reason to believe are less likely to be right, then suspension of judgment is appropriate.

Ultimately, McDonald et al. may disagree with me less than it appears. They do not conclude that “moral judgments are not justified” but only that “they are not justified by intuition alone.” My concern is only to examine empirical evidence that bears on the question of whether many of our moral beliefs are justified, whether by intuition or inference. As I emphasize in chapter 3, moral judgment involves a great deal of (unconscious) inference, even if the resulting moral judgments seem (consciously) to the individual to be automatic gut feelings.

R2.4. Disagreement about disagreement

In chapter 5, I argued that moral disagreement among epistemic peers should give most of us pause. Many controversial moral beliefs, of the average person at least, do not amount to knowledge and are less justified than they like to admit. This is one of the few skeptical challenges that I acknowledge is powerful, albeit limited in scope.

I focus my discussion on disagreement within a culture, partly because that is more tractable but also because I did not want to assume that moral truths are universal across cultures. D. Haas accepts my narrowed focus on disagreements among liberals and conservatives but argues that moral foundations theory cannot help identify foundational moral beliefs. One worry is that moral foundations merely express general values (e.g., loyalty, fairness), not beliefs in foundational moral statements, which are the stuff of disagreement and propositional knowledge (or lack thereof). It is true that the standard 30-item Moral Foundations Questionnaire (see Graham et al. Reference Graham, Haidt and Nosek2009, p. 1032) asks only, “When you decide whether something is right or wrong, to what extent are the following considerations relevant to your thinking?” The options include only general moral values, such as “Whether or not someone acted unfairly” and “Whether or not someone showed a lack of loyalty” (Graham et al. Reference Graham, Haidt and Nosek2009, Appendix A).

However, Haidt and his collaborators (Graham et al. Reference Graham, Haidt and Nosek2009, Study 2, Appendix B) have explicitly asked participants for their attitudes toward foundational moral propositions, such as:

• “It can never be right to kill a human being.”

• “People should not do things that are revolting to others, even if no one is harmed.”

Yet the researchers found similar results when asking participants to indicate their agreement with such moral statements (Graham et al. Reference Graham, Haidt and Nosek2009, pp. 1034–35). This suggests that the many studies employing foundational values are also tapping into foundational moral beliefs.

D. Haas also objects to my claim that liberals and conservatives share many foundational moral values/beliefs. Even if all moral foundations are recognized by most liberals and conservatives, differences remain among the groups’ relative weights of the values. If that is right, then there is more foundational moral disagreement than I admit, which would make the skeptical challenge affect many more moral beliefs. However, as I say in the book, differences in foundation weighting should not be overblown. Most people are not on the ends of the ideological spectrum, and yet those toward the middle express little variation in the foundational value.

In response, D. Haas argues that, although many liberals do value loyalty, authority, and sanctity, they do not “consciously acknowledge this and … typically will disown using these foundations.” That is true (for summary, see Graham et al. Reference Graham, Haidt, Koleva, Motyl, Iyer, Wojcik, Ditto, Devine and Plant2013, p. 96), but I suggest we take liberals’ intuitive responses as more representative of their fundamental moral values/beliefs. Compare implicit versus explicit racial biases. We should discount one's explicit disavowal of racial biases when one nevertheless exhibits them implicitly, especially given the motivation to appear unbiased. Similarly, moderate liberals want to appear clearly liberal, so they will be motivated to express greater commitment to foundations valued further on the left. (Perhaps we do not observe this effect as much in moderate conservatives because, although extreme conservatives value all foundations more equally, a motivation to appear more conservative need not involve discounting any particular foundations.). At any rate, often we can better determine a person's fundamental attitudes by investigating one's intuitive responses, because one's implicit reactions can be a better guide to their true feelings (see Cameron et al. Reference Cameron, Payne, Sinnott-Armstrong, Scheffer and Inzlicht2017).

In sum, the research on moral foundations does seem to provide evidence of people's foundational moral beliefs. And the differences between most (moderate) liberals and conservatives, although real, should not be overstated, which limits the scope of the skeptical challenge. Although I agree with D. Haas that the Moral Foundations Questionnaire is not perfect, I believe it is a good start – better at least than only armchair speculation about people's foundational moral beliefs.

McAuliffe & McCullough instead take issue with my claim that the average person should regard many of their opponents as epistemic peers. They object that social-scientific research can show that, in the aggregate at least, some opponents tend to be epistemic inferiors – people who are less likely to have morally reasonable or correct views. If that is right, I may be conceding too much to skeptics, which would make me overly pessimistic about controversial moral judgments.

The research McAuliffe & McCullough cite is indeed useful but limited. Many of the studies concern reflection, which earlier I noted is only one kind of reasoning that does not always lead to morally sound views (see the response to Landy in sect. R2.1). More reflective does not always mean more rational. Other research cited by McAuliffe & McCullough concerns general intelligence. But again I would caution that some measures of intelligence, such as mathematical reasoning and “book smarts,” do not always yield a virtuous person. Moreover, as McAuliffe & McCullough make clear, there are many aspects of morality and rationality, but that makes it difficult to know whether an opponent lacks the rationality relevant to the controversial issue at hand. A degree in medicine or the ability to quickly solve a Rubik's cube will not necessarily make one an expert on the ethics of international trade agreements.

Another concern of McAuliffe & McCullough’s is that we should be able to identify epistemic inferiors if there is to be moral progress. They mention that abolitionists provided cogent arguments that were met with self-serving and empirically spurious counter-arguments in defense of slavery. Shouldn't we be able to identify moral trailblazers and disregard defenders of the status quo as epistemic inferiors?

I believe some of us can. As I say in chapter 5, we can successfully debunk particular kinds of moral beliefs, just not large swathes of moral cognition. But it is not easy to identify epistemic inferiors – whether individuals or groups – and doing so definitely is not just a matter of “counting the number of learned people who hold a certain moral point of view.” Indeed, I concur with McAuliffe & McCullough that one must “examine the reasoning and evidence that each side has.” When I say that most people lack some moral knowledge about controversial moral issues, I am assuming that most people form their controversial moral beliefs without engaging in an honest and thorough examination of the opposition. That is part of the path to moral knowledge amid controversy. The denial of knowledge at one time does not preclude gaining it at another (cf. McDonald et al. on acquiring “independent confirmation”). My position here is not far from Greene's (2013), who councils us to be cautious and reflective when it comes to forming moral beliefs about controversial issues. The difference is that I do not believe the reflective examination of moral arguments relies preferentially on utilitarian considerations or otherwise requires abandoning our automatic moral intuitions.

Zheng raises a related issue about how disagreement affects moral progress. She helpfully points out that it is often precisely amid moral controversies that we should “stand up firmly” in support of our beliefs. Zheng also points out that we should often defer to marginalized groups, for they are epistemic superiors regarding their own oppression. This is an important point, but I can concede it as far as it goes. My claim is only that many ordinary people lack knowledge about controversial moral issues on which reasonable people can disagree. But knowledge is not required for all actions, such as voting to abolish slavery, listening to the concerns of those in a minority group, and protesting the inhumane treatment of oneself or one's group. Indeed, greater intellectual humility can yield deference and open-mindedness.

Of course, if such open-mindedness is required of those in the oppressed group too, then should they no longer confidently assert their grievances? Not necessarily. First, even in the absence of moral knowledge I admit a good deal of moral justification, which may be sufficient for action (cf. Locke Reference Locke2015). Even if we should only act on what we know (Hawthorne & Stanley Reference Hawthorne and Stanley2008), marginalized people do know what their experiences are. Although reasonable people might disagree, say, about whether a woman with white parents should identify as black, members of the black community know their concerns and should voice them (and others should listen). Moreover, a lack of knowledge about controversial issues need not paralyze one politically. One can, for example, exercise one's right to vote. In my view, voting is precisely about expressing one's concerns by backing the candidates or ballot initiatives one favors, regardless of whether one should be confident about their preferences.

In sum, my claim is not that most people lack any justification for their controversial moral beliefs; it is that they cannot claim to know. That is compatible with a good deal of justification for belief and for relevant action. My position is just incompatible with the kind of excessive confidence and intellectual arrogance one often sees in individuals.

R3.1. The primacy of emotion?

A common complaint is that I only target extreme, rather than sophisticated, forms of sentimentalism. On more modest versions (that at least still make empirical claims), even if emotions do not cause every moral judgment, they ultimately explain the moral judgments most people make. Kauppinen argues that there is a “striking parallel” between our “adaptive emotional tendencies and widespread patterns of moral judgment.” Kurth argues that emotions are to moral judgments as color experiences are to judgments of color. The mere feeling associated with shame, for example, “shapes and constrains” our concept of shame and its associated thoughts or content. Similarly, Clark & Winegard maintain that ultimately “we must build our moral judgments and arguments from the raw materials of our moral intuitions,” which they conceive of as “pre-rational preferences.”

In chapter 2, I do argue against such forms of sentimentalism by pointing to evidence that our emotional responses are typically consequences of our moral judgments, not vice versa. Kurth responds that the “experiments focused on just one emotion (disgust),” although for the record I also mentioned studies on compassion (p. 38). Kurth does question the value of one of the studies I draw on (Yang et al. Reference Yang, Yan, Luo, Li, Zhang, Tian and Zhang2013). The worry is that the experimenters only measured judgments of disgust, not feelings of disgust. But presumably the former typically relies on the latter. One typically judges something to be disgusting by feeling disgust toward it, or at least being disposed to feel that way – thus expressing a sentiment toward it (Prinz Reference Prinz2007).

Kurth and Kauppinen also raise a challenge for rationalism. As Kauppinen puts it, what a coincidence if “reason just happens to tell us to disapprove of the very things we in any case tend to feel negatively about.” Sophisticated sentimentalism nicely explains this striking parallel, for it is no surprise that we tend to praise acts that promote what we antecedent care deeply about and condemn acts that do not.

However, there is no coincidence if moral reasoning takes as its materials the things we antecedently care about. Imagine, for example, that utilitarianism is true and we can grasp the truth of the principle of utility by reason alone (Singer Reference Singer2005). Then moral reasoning will take as input what people care about, namely what makes them happy and sad. The same goes for other prominent moral theories. On Kantian rationalism, for example, we take our antecedent goals, plans, or maxims and test them against morality's demand that they be universalizable (Korsgaard Reference Korsgaard1996a). On contractualism, we similarly test existing plans to see whether they can be reasonably rejected (Scanlon Reference Scanlon1998). On such forms of rationalism, it is no mere coincidence that we happen to tend to judge wrong what we antecedently dislike. Genuine moral judgment makes heavy use of these starting materials in reasoning to moral conclusions.

Indeed, rationalism nicely explains why we do not always judge to be moral that which promotes our given desires, preferences, or goals – because only some of them are rational. Of course, rationalists must then maintain that our preferences are not themselves moral judgments. Clark & Winegard seem to think otherwise: “Moral judgments, by their very nature, must be grounded in intuitions about what is good and what is bad.” But judgments of good and bad are not sufficient for moral judgment. As Nichols (Reference Nichols2004, p. 15) has pointed out, we do not prefer natural disasters, and we think they are bad, but we do not judge them to be immoral.

R3.2. The structure of moral reasoning

Setting aside my critique of sentimentalism, some commentators raised challenges for my account of moral reasoning. That account was certainly under-described. More specific models of reasoning generally might help to illuminate moral cognition, such as the erotetic theory discussed by Alfano (based on work by Philipp Koralus). (Indeed, I suspect the erotetic theory provides a useful analysis of what it is to form a belief “on the basis of” another belief.) However, the already broad scope of the book demands certain limits on which issues I can address (more on that in sect. R4).

Take Kauppinen’s example of a simple moral argument: “Clinton lied. Lying is wrong. So Clinton did something wrong.” How, Kauppinen asks, do we reason to the moral principle that lying is wrong? In the book, I point to more fundamental moral principles, such as the principle of agential involvement (p. 69), but Marshall rightly argues that this cannot be the final foundation given its “all else being equal” clause. Marshall predicts that I would not want to posit a single fundamental moral principle, such as Kant's categorical imperative or the principle of utility, that is devoid of exceptions. I am actually open to that possibility. But my aim has only been to argue for rule-based moral reasoning that need not rely on emotions (conceived as distinct from such reasoning). I try to remain neutral on what the (rationalist-friendly) foundations of moral reasoning are, largely because I do not think we have enough empirical evidence to settle that psychological question.

With that said, let me say a bit more about what I take to be the options for rationalism. Now, the rationalist thesis at issue here is merely psychological, not normative, but moral epistemology provides various prescriptive theories that can also serve as psychological models (for review, see Zimmerman Reference Zimmerman2010). On infinitist theories, moral reasoning can just go on indefinitely, with the better justified moral views being the ones that are supported by more and better reasons. On coherentist theories, moral reasoning involves developing an internally consistent set of beliefs, perhaps through the venerable method of reflective equilibrium (compare Holyoak & Powell). On foundationalist theories, moral reasoning bottoms out in non-inferential states, such as moral perceptions or self-evident moral truths.

Which of these views most accurately represents human moral reasoning? I am not sure; we currently lack enough evidence. But all of the models are rationalist-friendly. Now, each model must ultimately address a foundational question: What exactly makes a belief distinctively moral? (Infinitists, for example, still need to determine when one has a moral reason.) That is another question I will not attempt to fully answer, because it is not necessary for carving out options for empirically sound rationalism. The rationalist need only deny that a distinctively moral judgment requires emotions (conceived as distinct from reason), which is precisely the aim of chapter 2. Once that is in place, the rationalist can adopt a coherentist, foundationalist, or infinitist model of belief formation in ethics (or even some combination of these).

Of course, there are sentimentalist versions of these models. For example, drawing on the analogy with color perception, sentimentalists might argue that the non-inferential foundations of moral reasoning are moral perceptions that are necessarily emotional in nature (compare Clark & Winegard; Kauppinen; Kurth). But my challenge to these and other sentimentalist views is to explain the sorts of issues taken up in the previous section, such as: (a) Why don't incidental emotions generally seem to change moral judgments (even if they need not always do so according to sentimentalists)? (b) Why do emotions often seem to be the consequences of, or dependent on, our prior moral judgments?

J. Haas raises a rather different concern about my view of moral reasoning. She argues that, instead of rule-based inference, we can explain moral judgment in terms of domain-general valuation mechanisms familiar from work on reinforcement learning. These mechanisms are certainly important and are beginning to contribute insights into moral psychology. I omitted discussion of them in the book partly for reasons of space, so I appreciate the opportunity here to say why at this stage I do not find them as useful for understanding moral judgment.

Consider how J. Haas proposes to account for the relevance of harmful outcomes to moral cognition. Such outcomes can be represented by multiple domain-general valuation mechanisms, but just consider the model-based mechanism which, as she defines it, “explicitly represents possible choices and determines the sequence of actions that maximizes value.” However, such a mechanism just models which outcomes promote what the individual takes to be of value, and even consequentialists agree that representations of value do not suffice for moral judgment. Consider two choices that generate a morally valuable state of affairs because they promote happiness. On a utilitarian view, only the one that brings about the most happiness will be the right thing to do – a deontic, not merely evaluative, status (cf. Mikhail Reference Mikhail2011, p. 85). A model-based mechanism can only account for moral judgment if it factors in that deontic step. Yet I have suggested that assigning the deontic status to an act involves categorization through the application of a moral rule or rules (the specific content of which I remain neutral on). We probably do not have enough evidence to settle this dispute, but until then I believe J. Haas's proposal is incomplete.

Let me say briefly why my bet is on a rule-based explanation. Domain-general valuation mechanisms may often appear to appropriately model moral cognition because, as Kauppinen and Kurth point out, there is that striking parallel (though not equivalence) between the things we judge right/wrong and the things we like/dislike, value/disvalue. Key to moral judgment, however, is not only its deontic status, but also its applicability to third-parties that have no relation to one's own personal choices. For example, I make a paradigm moral judgment when I read about atrocities abroad (or in fiction) and think “That's just wrong” – despite there being no connection to a choice I make to achieve some personal goal. I take it that, although many animals have evaluative attitudes, they lack this sort of core moral judgment. Yet the human brain shares domain-general valuation mechanisms with many other animal brains. Indeed, much of the work on reinforcement learning involves experiments with rats and monkeys in order to model the non-moral decisions they make to personally acquire what they value, such as food. For at least this reason, I suspect the relevant models of moral judgment will always be incomplete. The relevant mechanisms are developed to explain personal decisions underwritten by attitudes with only evaluative, not deontic, content.

Of course, we may eventually find that domain-general valuation mechanisms provide a complete explanation of core moral judgment. That is not necessarily a problem for rationalism anyhow. Even if the valuation mechanisms all rely ultimately on valenced flashes of affect (cf. Railton Reference Railton2017), they do not conflict with the principal rationalist idea that moral cognition is continuous with other forms of cognition and reasoning, that there is no special moral module whose operations are fundamentally different from reasoning generally (cf. Ch. 1, p. 11).

R3.3. Slaves of the passions?

Reasoning may generate new moral beliefs, but can it generate new motives or desires? As a Humean about motivation, Sinhababu is unconvinced; reasoning, he contends, only tells us how to satisfy our antecedent desires. One of his concerns is that I place too much weight on our ordinary explanations of action that do not appear to always posit antecedent desires, such as: Yongming chose salad over pizza because he thinks the carbs will make him too sleepy on the long drive home. Such explanations occur in everyday communication, which often omits the obvious, including background desires. Surely the explanation of Yongming's temperance, for example, tacitly assumes a desire to stay alert while driving.

Fair enough, but my reliance on ordinary explanations was merely meant to do dialectical duty. My aim was to demonstrate how anti-Humean explanations can work and to shift the burden of proof onto Humeans who must always posit an antecedent desire. The examples I draw on also help to illustrate how the Humean explanation is not any simpler, for it requires positing that extra antecedent desire.

Another key contention is precisely such claims about parsimony. Sinhababu maintains that the Humean theory is simpler because it posits only one causal relation for motivation, one in which desires cause beliefs, never the other way around. In the book, I suggest that the anti-Humean theory does not posit an additional causal process, because we should carve up the causal relation more broadly – whether the cause is a desire or a belief, it is mental causation among propositional attitudes. For comparison, I give the example of X baked Y, which remains the same process or relation, even when the “relata” (X and Y) vary. Sinhababu replies that we should individuate a process by its relata, at least when they are “essential to characterizing the processes.” But why should we treat desires or beliefs as essential to the causal processes in this debate? Sinhababu answers that otherwise Humeans and anti-Humeans cannot have a dispute in the first place. By comparison, he suggests that we should treat reasoning and telepathy as different processes in order to make sense of disputes about the existence of telepathy.

However, to recognize the differences between our two theories of human motivation, we do not need to treat the different causes of desires as distinct processes. We need only recognize their different implications for ethics (e.g., whether beliefs can ever motivate). To illustrate, go back to the baking example. Suppose we have two different theories. One says both humans and robots bake, whereas the other says only humans bake. Imagine further there is an ethical upshot: if robots also bake, then more humans will soon be out of a job. This empirical dispute remains, even if we assume baking is baking whether it is done by a human or a robot. Indeed, it would be odd to suggest that in order to have a dispute here we would have to posit two different types of processes: human-baking and robot-baking. (Such a move reminds me of those opposed to marriage equality who insist on distinguishing marriage from civil unions because of the relata in the relationship.)

In sum, I concede that these issues are difficult to settle, and I do not pretend to have proven definitively that we are not ultimately “slaves of the passions.” In chapter 8, I am largely on defense, aiming to show that when it comes to theories of human motivation, the Humean theory is not the only empirically sound game in town.

R4.1. The moral domain

Despite being concerned with distinctively moral judgment, I never provide a complete analysis of morality itself. As usual, however, if possible I avoid taking a stance on such controversial issues. My account is quite compatible with a number of specific views about the content and contours of moral norms – e.g., whether they concern mutual recognition (Carpendale & Wallbridge), the Kantian construction of one's practical identity (Roberts), or universalization of one's maxims (Marshall). On all of these specific proposals, moral judgment can be fundamentally a matter of reasoning.

Other commentators worry that such omissions have dire consequences. Chater, Zeitoun, & Melkonyan (Chater et al.) rightly stress that moral reasoners are “not lone and omnipotent decision makers” but embedded in social groups (cf. also Zheng). Yet most of the research I draw on measures individual moral judgments about dilemmas in which all relevant information is purportedly present. Moreover, I focus on mature moral thought and action among adults, without delving greatly into moral development and cross-cultural variation (Carpendale & Wallbridge; Narvaez; Rottman). As Narvaez puts it, “a wider examination of human behavior across time and societies is needed when discussing human moral potential.”

To be fair, I do discuss children's development of altruistic motivation to help others (Ch. 6). I also discuss moral knowledge among groups, particularly the masses (Ch. 5), as well as the social nature of moral cognition that lies in our evolutionary past (Ch. 4). Moreover, my position is supported by some appeal to cross-cultural research (see, e.g., Ch. 3). Although these discussions are all too brief, my aims were largely to demonstrate that optimistic rationalism is empirically defensible. To do that, I did not have to cover all of the scientific evidence or show that the picture I sketch perfectly describes all human beings and all the ways they interact. Indeed, I deliberately avoided committing to such claims about universality or innateness (see, e.g., Ch. 3, pp. 77–78).

R4.2. Moral truth and objectivity

Gibson worries about how my optimism is restricted to merely justified belief and appropriate moral motivation. A truly anti-skeptical defense should “provide a picture according to which thought and action can be meaningfully connected to the normative.” Whether a moral judgment amounts to knowledge, for example, depends not only on whether it is justified, but also whether it is true. An ancient Greek may be justified in believing that slavery is ethical, all while being ultimately motivated to uphold the practice for morally relevant (even if empirically misguided) reasons, such as justice and fairness. If that is the status of the moral mind, how can we be optimistic?

It certainly is part of a complete anti-skeptical project to explain how we are in touch with moral reality (cf. Marshall Reference Marshall2018). But it is possible for justification to have some connection to reality. If we adopt a reliabilist picture, one's moral beliefs are only justified if they are reliably accurate. I remained neutral on such disputes about the nature of justification, but this shows how justification and reality are not necessarily separable. And a similar approach can be taken to virtuous motivation.

In any event, I set aside such metaphysical issues in the book because most of the “empirical pessimists” I engage with do. Evolutionary debunkers, for example, often argue that, even if our moral beliefs are true, they are only accidentally true and so unjustified (e.g., Joyce Reference Joyce2006). Similarly, sentimentalists are “pessimists” about reason at least, but they are not defending a disconnect between moral judgment and reality. Quite the opposite: They believe emotions are precisely what put us in touch with such reality, even if that reality is something we ultimately construct (e.g., Prinz Reference Prinz2007). In fact, let me emphasize that (contra Carpendale & Wallbridge) I neither assume nor endorse cultural relativism or any theory which takes moral truths to be subjective. My aim was to remain neutral about such issues in moral metaphysics because I did not need to take a stance on them to address my topics of justification and proper motivation.

R4.3. On the armchair

The book focuses heavily on empirical evidence, yet some important philosophical arguments in moral psychology are decidedly non-empirical. Kauppinen even contends that “what defines the various sentimentalist views are the conclusions of the a priori arguments.” However, my aim of course is only to address versions of sentimentalism that make empirical claims about human psychology, not any theory worthy of the label “sentimentalism.”

In the end, the book is ultimately “on the pessimist's terms,” as Roberts nicely puts it, which are empirical. Rationalists too, though, rely sometimes extensively or exclusively on non-empirical premises to defend their views. Roberts points out that some Kantians, such as Christine Korsgaard, would insist that there is a “deep, perspectival divide between psychology and ethics.” We cannot settle the issue here, but I am of course dubious of philosophical views that make assumptions about human psychology – even just its possibilities and limits – and yet take empirical results to be irrelevant. Roberts's comments also make clear that optimistic rationalism is not a specifically Kantian view, let alone a version that posits an unbridgeable gap between ethics and empirical psychology. I hope the book, as well as the insightful commentaries in this journal issue show how to at least start building that bridge while respecting the philosophical terrain.