2.1. Emotion

A multitude of studies seemingly suggest that emotions alone affect moral judgment, not merely because they can affect inference by, say, directing our attention. I start with reconsidering the popular studies (Ch. 2), before going on to adduce evidence of moral inference (Ch. 3).

There are three main lines of evidence in favor of sentimentalism, and most of the evidence focuses on the emotion of disgust. First, feelings may seem necessary for conceiving of a norm as distinctively moral rather than a mere convention. For example, the norm against sexual harassment at work seems a matter of ethics, whereas the norm against wearing pajamas to work a mere matter of social propriety. Shaun Nichols (Reference Nichols2002) has argued that we treat moral norms as distinctive partly because we have strong feelings toward violations of them. However, there is far too much weight placed on the moral/conventional distinction as diagnostic of moral judgment. Even if people rate a norm as slightly more like a convention when they lack strong feelings toward it, that is not enough to demonstrate that norms are genuinely moral only if we have such feelings. Moreover, the relevant studies fail to manipulate emotions as a variable and are difficult to replicate in some circumstances (see, e.g., Royzman et al. Reference Royzman, Leeman and Baron2009).

Second, sentimentalists have drawn on studies in which participants’ manipulated emotions seem to cause changes in moral judgment (see, e.g., Prinz Reference Prinz2007; Sinhababu Reference Sinhababu2017). Famously, for example, participants inhaling a foul smell apparently think incest is morally worse than do participants in a control group (Schnall et al. Reference Schnall, Haidt, Clore and Jordan2008). There are many reasons why such studies, although numerous, fail to support sentimentalism (May Reference May2014). The main problem is that meta-analyses suggest the effects are tiny, perhaps even non-existent (Landy & Goodwin Reference Landy and Goodwin2015). Both the control and manipulation groups, for example, tend to rate the morality of the target actions the same. The mean differences, when found, are miniscule. Statistically significant does not mean significantly different (in the ordinary sense of the word). Now, there is a burden on the rationalist to explain why incidental emotions could ever have an effect on moral judgment, even if rare and ever so slight (Prinz Reference Prinz and Matthew Liao2016). But I provide an explanation (in Ch. 2) in terms of our well-known susceptibility toward misattributing the causes of our feelings (see, e.g., Schwarz & Clore Reference Schwarz and Clore1983).

Finally, emotions may seem essential to moral judgment because dysfunction in “emotion areas” of the brain seem to lead to moral incompetence (see, e.g., Nichols Reference Nichols2004; Prinz Reference Prinz2007). Psychopathy is the prime example (although I also discuss so-called “acquired sociopathy” and frontotemporal dementia). Psychopaths are characteristically callous, manipulative, and deficient in guilt and compassion (Glenn & Raine Reference Glenn and Raine2014). Some studies suggest that people with psychopathic tendencies somewhat struggle to draw the moral/conventional distinction (see, e.g., Aharoni et al. Reference Aharoni, Sinnott-Armstrong and Kiehl2012), but it is doubtful that this is enough to attribute significant deficits in moral judgment to them. Moreover, it is often underappreciated that psychopaths exhibit not only “emotional” deficits, but also clearly rational or inferential ones. Psychopaths are notoriously irrational, particularly imprudent, as a result of their poor attention spans, impulsivity, difficulties learning from punishment, trouble detecting emotions in others, and so on (see, e.g., Maibom Reference Maibom2005; Marsh & Blair Reference Marsh and Blair2008). I conclude that, although psychopaths likely exhibit some deficits in moral judgment, these should not be overstated (compared to their deficits in moral motivation) and that the moral ineptitude they do exhibit can be explained in terms of their deficits in reasoning.

Another problem with the appeal to psychopathology arises from a broader concern about the supposed reason/emotion dichotomy. Talk of “emotion areas” of the brain has become rather dubious in light of evidence that functionally diverse brain networks, extended over clusters of brain areas, give rise to emotions and other similarly complex psychological phenomena. Partly for this reason, emotional processing appears to involve a great deal of unconscious inference, involving the application of concepts, categories, and prior knowledge. So, for example, psychopaths suffer from dysfunction at least in the amygdala and ventromedial prefrontal cortex, but these areas are part of networks that facilitate not only emotion, but also unconscious learning and inference more generally (see, e.g., Woodward Reference Woodward and Matthew Liao2016).

The reason/emotion dichotomy begins to look rather spurious, as many philosophers and scientists are starting to recognize (see, e.g., Huebner Reference Huebner2015). But this does not mean the rationalism/sentimentalism debate is confused or pointless. What we are learning is that emotions involve a great deal of inference or, to put it the other way around, that inference is infused with affect (cf. Railton Reference Railton2017). This realization roundly supports the rationalist view that feelings are not required for distinctively moral cognition. Rather, moral cognition is like other forms of cognition: it requires unconscious inference that is facilitated by feelings or affect. This does not sit well with the sentimentalist tradition, which maintains that moral judgment, with its need for emotions, is importantly different from other domains of cognition.

Moreover, the affect that underwrites inference is a mere twinge of feeling, not traditional moral emotions, such as guilt, indignation, and compassion. Although such emotions are undoubtedly a prominent character in the drama of moral life, it is often because they are the normal consequences of our moral beliefs. For example, people who are vegetarians for moral reasons are more likely to become disgusted by meat (cf. Rozin et al. Reference Rozin, Markwith and Stoess1997). Compassion is likewise modulated by prior moral judgments. For instance, people feel little compassion for a student who missed classes because she left town with friends, but they readily sympathize with a student who missed classes because she was involved in a car accident (Betancourt Reference Betancourt1990). Similarly, those of us who react so passionately to racism, misogyny, and mass shootings do so because we believe they are terribly wrong. And we believe these acts are terribly wrong because we reason – we recognize, we learn, we infer – that they involve egregious violations of norms, which prohibit intentionally or recklessly causing unwarranted harm, disrespect, and so forth.

2.2. Moral inference

Let us now turn briefly to the positive case for moral reasoning. The crucial move here (Ch. 3) is to recognize that reasoning can be, and often is, unconscious. We can stipulate that the term “reasoning” only refers to conscious reasoning, but that is overly restrictive and unhelpful (Arpaly Reference Arpaly2003; Mallon & Nichols Reference Mallon, Nichols and Doris2010). Indeed, one of the counter-intuitive lessons from decades of convergent results in experimental psychology is that much of one's mental life is unconscious. That includes reasoning or inference, in which we form new beliefs on the basis of previous ones. For example, think about when you watch the opening scenes of a film – even a kids’ movie – which typically leaves important information implicit, such as the relationships among characters. Viewers are often left to infer what is going on, but it is not as though we consciously go through all the steps – “Ah, they look to be living in the same dwelling, yet they are in separate rooms and they both look sad, exhausted, and angry. Ergo, they must be in a romantic relationship and just had a fight!” Even if you could reconstruct something like this reasoning, it need not have been conscious at the time.

Moral cognition is no different. There is now a rather extensive scientific literature which reveals that intuitive moral judgments are driven by largely automatic and unconscious inferences, particularly about the consequences of the agent's action and how involved the agent was in bringing them about. “Agential involvement” turns on well-known distinctions in moral philosophy between acts versus omissions, intentional versus accidental actions, and harming as a means versus as a side-effect. Much of this literature employs the infamous trolley cases, but many of the studies ask participants to make moral judgments about more realistic scenarios. Besides, these hypotheticals have been useful for probing automatic moral intuitions across the globe and revealing that they are shaped by a variety of unconscious inferences about how much harm the action caused, whether it was intentional, whether it was an action versus an omission, and so on (see, e.g., Barrett et al. Reference Barrett, Bolyanatz, Crittenden, Fessler, Fitzpatrick, Gurven, Henrich, Kanovsky, Kushnick, Pisor, Scelza, Stich, von Rueden, Zhao and Laurence2016; Cushman et al. Reference Cushman, Young and Hauser2006; Young & Tsoi Reference Young and Tsoi2013).

Perhaps the most contentious corner of this literature involves the distinction between bringing about an outcome as a mere byproduct of one's action as opposed to a means to one's end goal. Some studies have failed to replicate early demonstrations of this means/byproduct effect, which is a core element of the old Doctrine of Double Effect. However, drawing on a recent meta-analysis of more than 100 studies involving more than 24,000 participants (Feltz & May Reference Feltz and May2017), I conclude that the means/byproduct effect is a real, even if small, aspect of agential involvement.

Moral inference is not always unconscious of course. I distance my account from extreme versions of the “linguistic analogy” or moral grammar hypothesis (Mikhail Reference Mikhail2011), which posit an innate moral faculty that is highly modular and impervious to conscious reasoning. I adopt an extremely minimalist dual process account (cf. Campbell & Kumar Reference Campbell and Kumar2012), on which moral cognition can be generated by both slow, conscious thought and automatic, unconscious processes. But there is no sound empirical reason to cast either mode of moral thought as uniquely unreliable, driven by emotion, or even “utilitarian.”

Throughout the book, I attempt what might be an impossible task: remaining neutral on what emotions are exactly. An ecumenical approach is enough, however, to generate a problem for sentimentalists. Suppose I come to realize that my country ought to take in Syrian refugees, but only after watching a video of the crisis. The video generates intense compassion that focuses my attention on their suffering which previously I had not fully recognized. Such emotions are relevant only insofar as they contain or cause changes in patterns of inference, attention, recognition, and the like. So, if emotions contain cognitive elements, then they can directly shape moral cognition by, say, directing one's attention and vividly highlighting morally relevant features of a situation. If emotions are mere feelings, lacking any cognitive elements, then they can only hope to shape moral cognition indirectly by changing patterns of inference. Either way, emotions can influence moral judgment in the way that they can influence any kind of judgment – by shaping inference through directing attention and so on. An unexpected mathematical claim, for example, can generate a feeling of surprise that directs my attention to new information and thus changes my inferences. Whatever emotions are exactly, they get a grip on moral cognition via reason and in a way that is not particular to distinctively moral cognition.

3.1. Defusing debunking arguments

Chapter 4 shows that wide-ranging skeptical arguments succumb to a Debunker's Dilemma (Kumar & May Reference May2018). Debunking arguments in ethics rely on an empirical and a normative premise (Kahane Reference Kahane2011; Nichols Reference Nichols2014):

1. 1. Some of one's moral beliefs are mainly based on a certain factor.

2. 2. That factor is morally irrelevant.

3. 3. So: The beliefs are unjustified.

But the two premises are difficult to jointly satisfy when one's target is large, because moral cognition is influenced by a variety of factors and these factors are only problematic in some contexts.

Take disgust. Although incidental feelings of this emotion are surely morally irrelevant (good normative premise), we have seen they hardly affect moral beliefs, if at all (bad empirical premise). Now, integral feelings of repugnance can influence moral cognition. Disgust toward the actions of sexists and corrupt politicians, for example, is typically tracking morally relevant information (cf. Kumar Reference Kumar2017). But a sound empirical premise is now joined with an awful normative premise: attuned emotions are not debunking.

Framing effects suffer the same fate. For example, the mere order in which information is presented is morally irrelevant (good normative premise), but meta-analyses (Demaree-Cotton Reference Demaree-Cotton2016) suggest that the vast majority of moral beliefs are unaffected by mere differences in order (bad empirical premise). Some moral beliefs might be substantially changed by mere framing (e.g., Tversky & Kahneman Reference Tversky and Kahneman1981), but meta-analyses suggest these are outliers (Kühberger Reference Kühberger1998), and wide-ranging critiques need trends.

What about Darwinian forces? Our moral beliefs are undoubtedly influenced by our evolutionary past. However, although mere evolutionary fitness is morally irrelevant (good normative premise), that is not a main basis for our moral views (bad empirical premise). The proximate causes of our particular moral judgments are values such as altruism, reciprocity, justice (which induces a desire for the punishment of norm violators), and so on. The ultimate explanation of these values may involve the fact that they were fitness-enhancing in the Pleistocene, but as proximate causes these values are morally relevant considerations. Evolutionary debunkers might deny that we can rely on any moral values to assess the normative premise, but that is self-defeating (Vavova Reference Vavova2015). If we cannot help ourselves to an independent evaluation of the normative premise in the debunking argument, then neither can the debunkers in defending it. Too often debunkers mistakenly think their task is merely to raise the possibility of moral error rather than demonstrate it empirically (see May Reference May2013b).

(Note: Many evolutionary debunking arguments target moral realism, specifically the objectivity of morality, which is not my topic. My concern is moral epistemology. I remain neutral on whether moral beliefs, when true, are objectively true or whether ordinary moral judgments presuppose as much.)

Finally, let us briefly examine automatic emotional heuristics. Are our non-utilitarian commitments unwarranted because they are “sensitive to morally irrelevant things, such as the distinction between pushing with one's hands and hitting a switch” (Greene Reference Greene2013, p. 328)? That is a fine normative premise, but the corresponding empirical premise is untenable. As Greene acknowledges, experiments demonstrate that our moral intuitions are not particularly sensitive to pushing alone, but rather pushing that is done with intent or as a means to an end (Feltz & May Reference Feltz and May2017; Greene et al. Reference Greene, Cushman, Stewart, Lowenberg, Nystrom and Cohen2009). Indeed, our non-utilitarian intuitions are generally sensitive to how involved the agent was in bringing about a bad outcome. Of course, utilitarians believe this is morally irrelevant, but that begs the question at issue in their debate with non-utilitarians. Greene (Reference Greene2014) also says our non-utilitarian intuitions are driven by rigid heuristics that are applied to moral problems with which the heuristics have “inadequate evolutionary, cultural, or personal experience” (p. 714). Again, a fine normative premise, but our best evidence reveals that moral intuitions are much more flexible, particularly during childhood, as they change over time in light of new information and recent cultural developments (see, e.g., Henrich Reference Henrich2015; Railton Reference Railton2017).

3.2. Selective debunking and moral disagreement

Although wide-ranging empirical critiques of moral cognition are flawed, more selective debunking arguments can succeed (Ch. 5). For example, one might point to empirical research on disgust and cognitive biases to debunk certain attitudes toward homosexuality, human cloning, and factory farming – particularly among a certain group of believers. There is not enough evidence at the moment, but the Debunker's Dilemma is unlikely to be a barrier.

Another form of selective debunking appeals to consistency reasoning (Kumar & Campbell Reference Kumar and Campbell2012). Empirical evidence can reveal that we maintain different verdicts about two similar moral issues for morally irrelevant reasons. It could turn out, for example, that most people believe that harming pets is morally objectionable, whereas factory farming is not, primarily because pets are cute. Similarly, although it is too wide-ranging to critique all non-utilitarian intuitions, we can all agree that it is morally irrelevant whether someone you can easily help is simply near or far away. Yet we could acquire rigorous empirical evidence that people tend to believe they lack an obligation to aid refugees in other countries primarily for this reason. Now, I am unsure that any of these particular debunking arguments would eventually succeed, at least for a sizeable group of believers. But the point is that empirical debunking can be done – if done properly, which will typically be selectively.

I take much more seriously a different form of empirical critique, which comes from moral disagreement. Philosophers have been extensively examining whether we really know something when it is disputed by “epistemic peers” – people one should regard as just as likely to be right about the topic (e.g., McGrath Reference McGrath and Shafer-Landau2008). But there has been little examination of the relevant empirical premise of the corresponding skeptical argument (cf. Vavova Reference Vavova2014, p. 304):

1. 1. In the face of peer disagreement about a claim, one does not know that claim.

2. 2. There is a lot of peer disagreement about foundational moral claims.

3. 3. So: We lack much moral knowledge.

Yet there is a wealth of empirical data on moral disagreements. To locate foundational disagreements, we might be tempted to go straight for cross-cultural research. However, it is more powerful to identify epistemic peers lurking within one's own culture.

Here I draw on Haidt's (Reference Haidt2012) famous moral foundations theory. Within a culture, liberals and conservatives apparently disagree about the relative importance of five (or so) fundamental values:

• Care/Harm

• Fairness/Cheating

• Loyalty/Betrayal

• Authority/Subversion

• Sanctity/Degradation

Does this provide support for the second premise in the skeptical argument from disagreement? Perhaps, but the critique will be – no surprise – limited. First, not everyone is an epistemic peer. But that is true only so far as it goes, and the empirical evidence does suggest that we should all be humbler about our cognitive abilities, especially on controversial topics in ethics. Second, and more importantly, disagreements about the foundations should not be overstated. Most people are not extreme liberals or conservatives, and as a result most people tend to recognize all five values. We just apply those values more to different topics (e.g., purity of the body vs. purity of the environment), depending on our other beliefs. Liberals and conservatives do apply different weightings to the five foundations, but among moderate liberals and conservatives the differences are a fairly small matter of degree (see Graham et al. Reference Graham, Haidt, Koleva, Motyl, Iyer, Wojcik, Ditto, Devine and Plant2013).

Ultimately, many people do probably lack moral knowledge as a result of peer disagreement. But this is restricted to particularly controversial moral issues. Many people do or should recognize that their most controversial moral beliefs are disputed by people who are just as likely to be right (or wrong for that matter). Here we may just have sufficient empirical evidence to challenge a selective set of moral beliefs, at least among the masses. Still, the overall picture of moral cognition is not pessimistic.

4.1. Egoism versus altruism

One reason for action that often conflicts with morality is self-interest. You should return the lost bracelet or harbor the refugee, not because it comes with a financial reward or will enhance your reputation, but because it is kind, fair, or just the right thing to do. But chapter 6 asks: Can we ever ultimately act on anything other than self-interest?

Most philosophers think so, but scientists often treat such an egoistic theory as a live empirical possibility. Fortunately, there are decades of rigorous experiments that back up the philosophers. C. Daniel Batson (Reference Batson2011), in particular, has shown that empathizing with another in distress, and hence feeling compassion, tends to increase helping rates, and not because such helpers want to gain rewards or avoid punishment. Moreover, experiments reveal that infants and toddlers help others they perceive to be in need, even when helping is not expected and requires the children to cease engaging in a fun activity (see, e.g., Warneken Reference Warneken2013). We can, of course, always cook up an egoistic explanation of the data, but it begins to look strained and rather implausible.

One might argue that none of this amounts to ordinary altruism, because empathy causes one to blur the distinction between oneself and the other. One is either in a sense helping oneself (egoism) or not quite helping a distinct other (non-altruism). Some theorists have proposed exactly this sort of account and it has some affinity with traditions that actively encourage such self-other merging, as in the concepts of no-self in Buddhism and agapeic love in Christianity (see, e.g., Cialdini et al. Reference Cialdini, Brown, Lewis, Luce and Neuberg1997; Flanagan Reference Flanagan2017; Johnston Reference Johnston2010).

The problem with these proposals is that they cannot make sense of the data. The empirical support for a self-other merging account is flawed, but more importantly there is a conceptual problem (May Reference May2011). When one helps another, there is a first-personal mode of presentation required to navigate the distinct bodies (cf. Perry Reference Perry1979). I cannot, for example, actively help another person while conceiving of the two of us as merely them (third-personal). I need to know which arms and legs I must move to save her. Even us smuggles in a first-personal reference to a self. So we ought to treat empathy as inducing a concern for others represented as distinct from oneself. Ordinary altruism is thus possible and even prevalent, given that empathy, and the compassion it engenders, are not uncommon.

4.2. Rationalization and moral integrity

Pessimists might accept the existence of genuine altruism but argue on empirical grounds that it is limited, restricted primarily to our kith and kin. When we interact with acquaintances or strangers, we might be primarily motivated by self-interest or otherwise the wrong reasons. However, chapter 7 covers ample evidence that people are quite frequently motivated by their moral beliefs. Oddly enough, the evidence comes from studies of bad behavior, particularly when we succumb to temptation.

Consider, for instance, the phenomenon of moral licensing. After doing something virtuous or affirming one's own good deeds, one sometimes justifies bending the rules a bit. For example, one study found that participants were a bit less honest and generous after conspicuously supporting environmentally friendly products, compared to a control group (Mazar & Zhong Reference Mazar and Zhong2010). There are many studies of such moral licensing (Blanken et al. Reference Blanken, van de Ven and Zeelenberg2015), and they are just one form of the familiar phenomenon of motivated reasoning or, more generally, rationalization (Kunda Reference Kunda1990).

We can also look to studies in which one will fudge the results of a fair coin flip in order to steer a benefit toward oneself (Batson et al. Reference Batson, Kobrynowicz, Dinnerstein, Kampf and Wilson1997). Importantly, participants in such studies tend to rate their actions as morally acceptable, just because there is a sense in which they did use a fair procedure (flipping the coin), despite fudging the results in their favor. Flipping the coin provides just enough wiggle room for many people to rationalize disobeying the results. Such bad behavior is motivated not merely by self-interest but by a concern to act in ways one can justify to oneself – that is, by one's moral beliefs (or, more broadly, normative beliefs).

Notice that this is not just rationalization of one's bad choice after it happens (post hoc), but rationalization before the action in order to justify performing it (what I call “ante hoc rationalization”). This should be recognizable in one's own life. People do not just behave badly because it is in their interest. When they could just think (probably unconsciously) “I am going to keep this lost $20, because I want the money,” they instead think something like: “I probably need the money more than the owner,” or “I have done more than my fair share of good deeds this week,” or even “I bet it is that sleazy banker's money, and he has got plenty.” These are thoughts that could potentially justify one's behavior, even if the reasoning is addled. (Indeed, even atrocities are rationalized, unfortunately.) After the rationalizing is done, one does not necessarily see oneself as doing anything morally objectionable (cf. Holton Reference Holton2009). One's actions are in line with one's moral beliefs – at least temporarily, since later on one may be cool, calm, and collected or otherwise see matters aright, at which point guilt sets it.

What these various studies reveal is the motivational power of moral beliefs. Our focus has been on bad behavior because the relevant studies concern temptation. But there is no reason to think that moral beliefs play any less of a role in motivating good behavior. We are normative creatures, most of whom care deeply about acting reasonably and justifiably, whether we end up doing what is right or wrong. We care ultimately about acting in particular ways, such as being fair, which we regard as right, but we also ultimately care about doing what is right as such.

When we do what is right, then, we are not ultimately motivated by self-interest alone but by considerations we deem to be morally relevant or genuine reasons, such as considerations of fairness, justice, benevolence, loyalty, honor, and even abstractly “the good” and “the right.” Like Hurka (Reference Hurka2014), I adopt a pluralistic approach on which all these sorts of considerations are the right kinds of reasons or concerns (whether they are construed, to use some philosophical jargon, “de dicto” or “de re”). I adopt some terminology from Batson (Reference Batson2016) and call any such concerns to do what is right moral integrity. This is a third intrinsic concern that we should add to human psychology – in addition to ultimately caring about one's own self-interest (egoism) and the well-being of others (altruism). Indeed, moral integrity is plausibly related to the trait of “moral identity,” which varies in the population, and can be enhanced or suppressed (Aquino & Reed Reference Aquino and Reed2002).

4.3. The autonomy of reason

At this point, a theorist inspired by David Hume might argue that our moral beliefs, even if products of reason, are ultimately under the direction of desire (e.g., Arpaly & Schroeder Reference Arpaly and Schroeder2014). Suppose, for example, that while on the bus you offer your seat to an elderly man standing in the aisle. A Humean might argue that you are only ultimately motivated to act because you happen to care about being respectful or about doing what is right. Do we have empirical reasons to always posit such antecedent desires which our moral beliefs serve? Does the scientific evidence show that reason is always a “slave to the passions”? These questions are taken up in chapter 8.

We certainly sometimes do what we believe is right because we are antecedently motivated to do what is right as such (“de dicto”) or to promote particular moral values, such as kindness, respect, and fairness (“de re”). But this need not always be the case. On a sophisticated anti-Humean view (May Reference May2013a), one is capable of being motivated to do something simply because one believes it is the right thing to do, even if one has a weak or non-existent desire to do it or to be moral. For example, someone who engages in discriminatory behavior can be motivated to stop simply by coming to believe it is the right thing to do, even with no changes to his antecedent goals or motives. This anti-Humean picture is, despite appearances, entirely compatible with the science.

Take neurological disorders, which some Humeans have used to support their view. Here I will just mention the example of damage to the ventromedial prefrontal cortex. Patients with such damage who develop so-called “acquired sociopathy” tend to have difficulty making appropriate social, moral, or prudential choices. These patients seem to retain knowledge of how to act but struggle to translate their general normative judgments into a decision and action in the moment (Damasio Reference Damasio1994/2005). Some philosophers believe these patients support the Humean thesis that moral (or otherwise normative) beliefs cannot motivate by themselves (cf. Roskies Reference Roskies2003; Schroeder et al. Reference Schroeder, Roskies, Nichols and Doris2010).

But that is based on a misunderstanding of the opposition. Of course one's moral beliefs do not always generate the corresponding desire, but when they do they need not rely on an antecedent desire. Instead, the necessary element could be, say, a lack of full understanding – for example, a patient believes she ought to thank the host, but she does not entirely appreciate that she is in the relevant circumstances (cf. Kennett & Fine Reference Kennett, Fine and Sinnott-Armstrong2008). Or it could be that the relevant brain dysfunction disrupts her virtuous dispositions to be motivated to do what she knows she ought to do. Indeed, far from being incompatible with anti-Humeanism, acquired sociopathy reveals that normally our moral beliefs do motivate but that this can break down in cases of pathology.

Other empirically minded Humeans contend that desires are ultimately necessary for all motivation because they provide the simplest explanation of action (e.g., Sinhababu Reference Sinhababu2017). In particular, desires are goal-directed states that are inherently motivational, direct one's attention to their objects, and cause pleasure when one anticipates satisfying them. These are characteristic features of desire that arise out of our intuitive folk psychology but also neuroscience, particularly our understanding of the brain's reward system (Schroeder Reference Schroeder2004). Thus, it may seem that moral beliefs cannot play this same role without either being desires or being an unnecessary additional posit in psychology.

However, I show that desires do not have a monopoly on these psychological properties. Indeed, the reward system provides a framework for understanding any mental state that treats an event as positive or “rewarding.” In this way, moral beliefs (e.g., “Smoking near children is wrong”) share much in common with desires (e.g., wanting to smoke away from children), compared to merely descriptive beliefs (e.g., “Secondhand smoke causes cancer”). Both moral beliefs and desires treat a state of affairs as valenced – as good/bad or desirable/undesirable. The two states are importantly different, however, in that only beliefs are assessable for truth and thus suited to playing an integral role in reasoning – the forming of new beliefs on the basis of previous ones. However, when a belief does contain normative content, it represents its object in a positive light and thus typically generates some desire for it.

Consider how anti-Humeanism nicely explains a particular example. Suppose your friend goes on a meditation retreat and comes to realize that he is kind of a jerk. In conversations with others, he tends to boast, redirect the conversation toward himself, and rarely ask about his interlocutor's problems or concerns. (Or imagine another moral failing you or a friend struggle to correct.) On the anti-Humean view, we can explain this kind of scenario in terms of two independent sources of intrinsic motivation. The jerk has an egoistic desire to feel good about himself and discuss his own problems. But he also believes it is important to be a good person, and recently has become thoroughly convinced that he has some relevant character flaws here. This moral conviction or belief – indeed, knowledge – generates a new desire in him to correct his behavior. Now, the Humean would insist on positing an antecedent desire to be moral, which this new belief serves, but I argue that we do not have any empirical reason to always do so. Reason is not destined to be a slave to the passions.

5.1. Self-interest returns

Despite the existence of genuine altruism, we may too often rationalize serving self-interest, perhaps in a self-deceived manner. Even when we do what is right, we may often do so because unconsciously we ultimately want to curry favor or avoid being socially ostracized. Although the science does demonstrate that we can be ultimately motivated by more than self-interest, there is some evidence that this is less common than we would like to admit (Batson Reference Batson2016). Virtuous motivation is threatened given that acting from self-interest is often the wrong kind of reason to do what is right.

Consider again studies of fairness. In some experiments, many participants only appear to be fair, by flipping a coin to determine who gets a reward, yet around 90% of the time the flip magically favors the participant. Clearly, there is some fiddling of the flip going on. Follow-up studies suggest the fiddlers do not just misremember whether they chose heads or tails. Instead, they are primarily motivated to avoid seeing themselves as immoral (Batson et al. Reference Batson, Thompson and Chen2002). In fact, fiddlers rate their behavior as moral, unlike those who do not flip at all.

Batson interprets this as “moral hypocrisy,” which he regards as a kind of egoistic motivation to look good to oneself. However, although seeking to appear moral to others is clearly egoistic, being ultimately motivated to look moral to oneself is just a concern to be moral. This is moral integrity even though, as with other forms of motivated reasoning, one's conception of morally good behavior is corrupted at the time of temptation. Moreover, only some participants fiddled the flip, and only when there was enough “wiggle room” that they could justify flipping the coin but then ignore the results. We hardly have evidence for the cynical conclusion that our moral choices are dominated by self-interest alone without a concern to be moral.

Similar issues arise with studies of dishonesty. When participants can get away with it, many will lie about how many arithmetic puzzles they solved, in order to earn more money from the experimenters (see, e.g., Mazar et al. Reference Mazar, Amir and Ariely2008). Interestingly, dishonesty is mitigated significantly, often nearly eliminated, when participants are reminded of moral standards (see Ariely Reference Ariely2012). In one case, for example, participants were first asked to write down as many of the Ten Commandments as they could recall. In another study, participants had to sign an honor code before they took a crack at the puzzles. Both interventions significantly reduced cheating.

Once again, we might be led to think that, when moral choices are available, egoism is rampant. However, as Ariely makes clear, the vast majority of people only cheat a little by claiming to have solved about 10% more of the puzzles than they did, and this dishonesty can be mitigated with moral reminders. Indeed, whether or not people cheat, the mechanism appears to be rationalization. One rationalizes cheating a little, for that is all one can justify to oneself. Some even rationalize not cheating at all by having one's attention drawn to one's considered moral beliefs. Either way, the proper motivation appears to be in play: People are primarily motivated by a concern to act in ways they can justify to themselves as morally acceptable, not merely by self-interest. If their ultimate concern were self-interest alone, they would not have worried themselves about the morality of their choices.

The Defeater Dilemma is evident here. It is a plausible normative premise that acting from self-interest is often the wrong kind of reason to act. But a careful look at the evidence suggests instead that people are quite motivated to be moral, and the corresponding normative premise is thereby implausible. We are not motivated by the wrong reasons if we are motivated to do what is right. There is no doubt that we are motivated by egoism as well, but we should not overstate its power and prevalence, and likewise we should not ignore the power and prevalence of moral integrity (even when it is a result of motivated reasoning). The same dilemma arises for the challenge from situationism.

5.2. Situational forces

Countless studies support the situationist thesis that we are often unconsciously motivated by surprising features of our circumstances, at least more often than we intuitively expect. In one study, for instance, about twice as many participants at a mall helped someone make change for a dollar when in front of a bakery or a coffee roasting company, compared to participants who had the opportunity to help in front of a store that was not emitting such pleasing aromas (Baron Reference Baron1997). Similarly, participants are much less likely to help someone apparently in need of serious help if there are other people nearby who are not helping (Latané & Nida Reference Latané and Nida1981). And, infamously, people make decisions about who to hire and even who to shoot, based partly on implicit biases against the person's race, gender, and other social categories (see, e.g., Bertrand & Mullainathan Reference Bertrand and Mullainathan2004; Payne Reference Payne2001). These are just a few examples of the relevant sorts of studies. Some may not survive the replication crisis, but enough will likely remain to suggest that people can be influenced unconsciously by features of their circumstances.

Many philosophers and scientists have taken such results to threaten the existence of traditional character traits (cf. Alfano Reference Alfano2013) or of certain conceptions of free will and moral responsibility (e.g., Vargas Reference Vargas, Clark, Kiverstein and Vierkant2013b). However, the more fundamental worry is that we are not motivated by the right reasons: Did I act primarily to help the person in need or because the pleasing smell of cookies put me in a good mood? As Dana Nelkin (Reference Nelkin2005) has put it when discussing the threat to free will: “the experiments challenge the idea that we can control our actions on the basis of good reasons” (p. 204). Thus, the situationist literature might seem to fund a wide-ranging critique of what motivates moral behavior.

The Defeater Dilemma remains an obstacle, however. Some situational forces do substantially influence morally relevant behavior, thus grounding a strong empirical premise in the skeptical argument. However, then the normative premise suffers. For example, meta-analyses suggest that circumstantial changes in mood do significantly impact helping behavior (e.g., Carlson et al. Reference Carlson, Charlin and Miller1988), but the vast majority of studies concern acts that are morally optional or supererogatory. There, I argue, mood is a morally relevant consideration: Your mood is an appropriate consideration, among others, when deciding whether to help a stranger make change for a dollar, pick up some papers someone dropped in a mall, and so on. If helping is morally optional, then whether you feel like helping is a relevant consideration. Perhaps it is inappropriate to only help because you feel like it, but it is a relevant consideration that may tip the scales in favor of acting.

Other studies do have confederates who appear to be in serious need. It is not morally optional to help someone who, for example, appears to have fallen off a ladder. But here, too, the effects are driven by morally relevant considerations. In-depth studies of group effects suggest that most participants do not help in the presence of bystanders because participants firmly believe that no help is really needed (cf. Latané & Nida Reference Latané and Nida1981; Miller Reference Miller2013). Such a belief is unwarranted, but what it concerns is morally relevant.

The same cannot be said of other factors, such as implicit racial biases and genuine framing effects, which are clearly morally irrelevant. Here we have a plausible normative premise for the skeptical argument, but its corresponding empirical premise becomes untenable. Our moral decisions are sometimes partly determined by the mere order in which information is presented. But again meta-analyses suggest that the vast majority of moral decisions remain the same in the face of genuine framing effects (Demaree-Cotton Reference Demaree-Cotton2016). Although some framing effects produce dramatic results, these are outliers (Kühberger Reference Kühberger1998).

Similarly, although implicit biases no doubt exist, recent meta-analyses suggest that their effects are quite small and do not predict much behavior (Forscher et al. Reference Forscher, Lai, Axt, Ebersole, Herman, Devine and Nosek2017; Greenwald et al. Reference Greenwald, Poehlman, Uhlmann and Banaji2009; Oswald et al. Reference Oswald, Mitchell, Blanton, Jaccard and Tetlock2013;). Importantly, and I cannot stress this enough, that does not mean implicit biases cannot explain large-scale problems in society. Indeed, implicit biases may add up to explain the powerful discrimination any one individual experiences as a result of slights from many people, however well-meaning these people are. But the evidence to date does not suggest that most ordinary people base many of their morally relevant decisions primarily on their implicit biases. Some do, for sure, but we are looking for trends in the data that can fund wide-ranging critiques. When a police officer does the right thing and decides not to shoot an unarmed teenager who is brandishing a toy gun, it is probably not primarily because the suspect is white – although that may play a minor role. At other times, when the child is black, the minor role race can play might sadly be just enough to yield a pulled trigger and a lost life, particularly in a high-pressure situation when a split-second decision is made (which is precisely when most implicit biases show up in the lab). But, again, our inquiry concerns a main basis for most people's moral and immoral behaviors.

The foregoing is just a sampling of the situationist literature, but you get the idea. When targeting a wide range of morally relevant behaviors, it is difficult to identify a single influence that is morally inappropriate in all or nearly all contexts. Our moral decisions are based on many factors, only some of which are a main basis for any one choice. Moreover, a single influence can be inappropriate in some contexts but appropriate in others. Mood is sometimes an appropriate consideration when deciding when to help, but not when the situation is dire. Even race can be a morally relevant consideration in some contexts (e.g., when justifying certain affirmative action policies). Thus, rather than picking apart a few studies among many, I aim for the Defeater Dilemma to provide a principled and systematic way to resist challenges to virtuous motivation from situationism and related literatures.