In a series of recent articles, psychologist Paul Bloom (, , ) and philosopher Jesse Prinz (, ) have attacked the many recent positive evaluations of empathy for moral functioning (e.g., de Waal ; Rifkin ). Although current common wisdom sees empathy as important for or even at the core of morality, Bloom and Prinz argue that this view is not only wrong, but the opposite may be true. Not only does empathy not promote the general well‐being of society, they claim; empathy may even be harmful to the pursuit of justice. Because of this, Bloom and Prinz argue that the tugs of empathy must be resisted so that moral principles may be served. Rather than being part of the good life, empathy is instead part of an evolutionary heritage that must be overcome.

The purpose of this article is to (1) critique the primary arguments given by Bloom and Prinz against empathy and (2) argue instead for the view that empathy is best understood as a virtue that plays an important but complicated role in the moral life. The argument proceeds in six parts. In the first section, I address basic definitional issues which complicate the discussion, and I endorse, for the purposes of this article, the widely used distinction between cognitive and emotional empathy, noting that they are logically distinct even if in practice they are often conjoined, perhaps necessarily so. In the second section, I summarize the arguments of Bloom and Prinz against empathy, and I note how both affirm the claim that empathy is a fixed trait and how their argument against empathy hinges on this claim. Although Bloom and Prinz appear to be targeting both cognitive and emotional empathy, the latter is their primary target, and their primary concern is the way that emotional empathy can improperly contribute to parochialism. The third section briefly examines several of the experiments Bloom and Prinz rely on to support their case against empathy. Two things are notable. First, while some of these experiments do provide evidence that empathic reactions can be induced to promote actions and judgments that are immoral, the experiments do not establish the fixed trait claim that lies at the heart of the argument against empathy. Second, an examination of these experiments reveals that not all of the experiments are directly testing empathy, and this raises methodological concerns. The fourth section examines several more recent studies consistent with the claim that empathy is not a fixed trait and that it is modifiable by intentional intervention. Taken jointly, the second and third sections demonstrate that the empirical premise of the argument against empathy is false, and as a result the moral argument against empathy is invalid. The fifth and sixth sections lay out a positive framework for understanding the role of empathy in the moral life. The fifth section briefly lays out a theory of what a virtue is in philosophical and psychological terms. I define a moral virtue as an active disposition (character trait) that (1) is learned over time, (2) integrates implicit and explicit processing, and (3) is a necessary component of full moral functioning. In the sixth and final section, I then make the case that empathy fits this definition, and for that reason, we should regard empathy as a virtue. Empathy is a not a sufficient condition for moral judgment and action, but it is a necessary condition for full moral functioning. That being said, empathy is only a virtue among the virtues, and I argue for a role for empathy that participates in but does not replace rational reflection.

WHAT ARE WE TALKING ABOUT WHEN WE TALK ABOUT EMPATHY?

Empathy is a term of relatively recent coinage, a neologism invented to be a translation of the German term Einfühling (Titchener , 21). Although now thought of as distinct from the concept denoted by “sympathy,” the latter term was used by Adam Smith and David Hume much in the way we think of “empathy” now, and their term “fellow‐feeling” captures an important slice of how we currently speak when we speak of empathy. That being said, Steven Pinker has argued that the popularity of empathy in recent decades is due in part to the conflation of empathy proper with what we now think of as “sympathy,” so that empathy is thought to imply a positive action towards others, presumably in a morally appropriate way (Pinker , 574). As we shall see, this is a concern in some of the scientific literature as well, but there are also other issues, and extensive exploration and experimental study has led to the need to be increasingly precise in the term's usage. C. Daniel Batson () has recently argued that the word “empathy” is now used in up to eight ways, and that clarity in both science and philosophy will in the future require careful distinction among the forms of “empathy” being considered.

For present purposes, I follow a simpler typology that is increasingly used in both the scientific and philosophical literature, one that makes a primary distinction between cognitive and emotional empathy. Cognitive empathy involves the capacity to think about the thoughts and feelings of others, often described as theory of mind, mental simulation, or mindreading (see, e.g., Carruthers and Smith ; Goldman ; Spaulding ). While there continues to be competition between differing accounts of theory of mind and mental simulation, we can for our purposes here ignore them. One distinction is however worth mentioning, that between “putting yourself in another's shoes” and more straightforwardly imagining what the other as other would think and do in a given situation. Both skills are important, but the latter captures more of the heart of what we think of as empathy, because “putting yourself in another's shoes” may help you imagine how you think you would feel in the other's situation. This is not identical to, and is indeed less taxing than, trying to imagine the situation of other as other, whose psychology and preferences may be quite distinct from one's own.

Such cognitive empathy is conceptually distinct from emotional empathy, which involves in some sense a homology of feeling and emotion between self and other. As emotional empathy is commonly conceived, however, there is considerable ambiguity, since it could mean any of the following:

• (1)

  experiencing the exact same emotion at the exact same strength as another;

• (2)

  experiencing the exact same emotion, but not necessarily at the same strength;

• (3)

  experiencing not necessarily the same emotion but the same valence of emotion (positive or negative); or

• (4)

  experiencing emotional arousal as the result of witnessing the emotional arousal of another, but not necessarily the same emotion, the same valence, or same strength.

While (4) seems far too weak a candidate to be properly labeled empathy, (1) seems too strong, even though it best meets the conceptual requirements of what empathy entails. The obvious problem with (1) is that there is good reason to think that it happens only rarely. If I witness an Olympic skier wipe out in an horribly awkward and painful fashion, I may wince in empathetic pain with the injuries I witness, but such empathetic pain is undoubtedly much less severe than that experienced by the skier. Options (2) and (3) seem therefore the more likely candidates, though existing literature rarely makes the distinction between the two as they relate to empathy conceptually. For purposes here, I will treat both options as plausible candidates for emotional empathy, even though (3), which requires only identical valence of emotion, is a comparatively weak criterion.

There is reason to think, however, that what might be called “empathy proper” requires both cognitive and emotional elements, and several scholars have made that claim (Simmons ; Masto ; Spezio ). A reason for this is to clearly distinguish empathy from the related phenomenon of emotional contagion, an early version of which occurs when the crying of one baby triggers the crying of other babies in the same room. In such a case, the babies share the same emotion, but it is difficult to think of this as empathy, since the babies lack the awareness that their crying is being shared. For an emotional reaction to be empathetic rather than simply shared, there has to be some awareness that the emotion is being shared, and this would require by definition cognitive empathy as well. So emotional empathy seems to entail some level of cognitive empathy, but is the reverse true? It seems plausible to think that one can simulate or imagine the thoughts and feelings of another without triggering similar emotions in oneself, but this may be problematic. Work on the relation of emotion and cognition strongly implies that normal cognition relies on emotional processing, and that deficits in the latter can impair the former (Damasio ; Thagard ). The work of Damasio and colleagues in particular demonstrates that such effects impact moral functioning, and more recent work suggests that emotional and cognitive empathic processing is intertwined (Anderson et al. ; Young et al. ; Strombach et al. ).

For present purposes, I will not take a strong stand on this latter issue. Logically, emotional and cognitive empathy are dissociable, even if in practice they are often intertwined, and so in most instances I will assume that they are linked. We may distinguish first between contagion and empathy, and then between weak emotional empathy and strong emotional empathy, both including cognitive empathy, with the former (weak emotional empathy) involving only the same valence, and the latter (strong emotional empathy) involving the same emotion, but not necessarily at the same strength as the individual empathized with.

ARGUMENTS AGAINST EMPATHY AND THE FIXED TRAIT ASSUMPTION

The attention to definitions at the outset is of some importance, because how empathy is properly defined plays an important role in the evaluation of the arguments by Bloom and Prinz and the scientific sources they cite. Bloom () recognizes the cognitive–emotional distinction, and the specific arguments he produces seem to be aimed mainly at the phenomenon of emotional empathy. Prinz (2011a) alternatively defines empathy as experience of another person's state or “a kind of feeling for an object and a feeling‐on‐behalf of an object” or putting oneself in another's shoes, and at one point he identifies emotional contagion as a form of empathy (Prinz , 211). Although some of Prinz's arguments concern cognitive empathy, his primary target also seems to be emotional empathy and the way that emotional empathy in some cases motivates amoral or immoral behavior. It is noteworthy as well that in the arguments both Bloom and Prinz put forth concerning emotional empathy, the focus is entirely on negative sensations and emotions rather than positive ones, for example, pain, suffering, and sorrow rather than joy, pleasure, or admiration.

Both Bloom and Prinz put forth a number of arguments against empathy, and the general thrust of Bloom's arguments in particular highlights the ways in which empathy induces partiality and may be prone to manipulation. Anecdotally, Bloom highlights several cases that, he argues, reveal the problem with our empathic reactions, notably the 1987 “baby Jessica” case and the 2005 Natalie Holloway case. Both are cases that drew significant and prolonged media coverage, the former involving a 1‐year‐old child trapped in a well and the latter the disappearance and likely murder of a recent high school graduate while on a trip to Aruba. In cases such as these, he argues, empathy induces us to identify with the plight of a single individual, even one that we do not know but whom we can, thanks to the media, readily identify with and who seems “like us.” In the case of baby Jessica, Bloom notes the large sums of money donated to the cause of helping her family, sums that were contributed not to the rescue of baby Jessica but simply to help her and her family. In the case of Natalie Holloway, Bloom contrasts the amount of media and public attention to Holloway to the attention given to the concurrent genocide occurring in the Darfur region of Sudan. In both cases, according to Bloom, there is an immoral misallocation of priorities as well as resources and attention. Rather than helping us focus on the good of the whole, Bloom argues that empathy induces us to be disproportionately concerned with the individual even when the whole suffers as a consequence.

A second argument that Bloom makes (both in 2013 and 2015) is that empathy is problematically prone to political manipulation. Although this argument is not well developed, it seems to be of some importance to Bloom, and he cites, again anecdotally, the surge of anti‐immigrant sentiment that accompanied the 2016 U.S. presidential race and the Trump candidacy in particular. The argument seems to be that candidates can play into and thus manipulate fears driven by the parochial character of cognitive and emotional empathy, that empathy encourages us to identify and bond with those “like us” at the expense of those who are “unlike us,” with the consequence that the latter are treated unfairly.

Like Bloom, the concern about the link between empathy and partiality plays a significant role in Prinz's argument that empathy is at best overrated and at worst a driver of immoral behavior. In addition, Prinz argues that, despite the view of many to the contrary, empathy is not a particularly powerful motivator of moral behavior, and that other emotions, such as anger, are better and more effective motivators. Both Bloom and Prinz make the argument that empathy is neither necessary nor sufficient for moral action. Empathy is not sufficient because the presence of empathy does not guarantee moral action. That empathy is not necessary, they argue, is demonstrated by the fact that individuals with Asperger's syndrome have impaired capacity for empathy but can nevertheless still function according to moral principles.

Both seem to imply that empathy is a fixed biological trait not amenable to modification, but neither Bloom nor Prinz are fully consistent on this point. Bloom () does argue that empathy is “bred in the bone” and notes that it can (must?) occur automatically and involuntarily. Prinz claims that empathy is intrinsically biased and states, “We can no more overcome its limits than we can ride a bicycle across the ocean; it is designed for local travel” (, 229). Despite this, both argue that our empathic reactions can somehow be overcome, and Bloom () states, “But empathy will have to yield to reason if humanity is to have a future.” Presumably, what they have in mind is the claim that, while we cannot control our empathetic reactions, we can control what we do with them, and what we should do with them is bury them as deeply as possible lest they interfere with our moral decision making. So, our empathetic reactions are fixed, but what we do with them is not. This notion of the fixity of the moral emotions is a view they share with moral sentimentalists like David Hume, although both disagree with Hume concerning the way in which emotions guide action. Although Prinz counts himself a moral sentimentalist, with respect to empathy both Prinz and Bloom's arguments are utilitarian in character, and it is noteworthy that their concerns about the partiality of empathy over and against the greater good have more than a little utilitarian air to them.

It is important to note the central role that fixity plays in their argument. If empathy is not a fixed trait, then the arguments by Bloom and Prinz against empathy make little sense: one would not need to be against empathy if our empathic reactions were subject to modification in beneficial ways. Neither Bloom nor Prinz argue that our empathic reactions are always morally negative. Rather, they argue that they are negative in certain and perhaps most morally relevant instances, the effects of empathy are rarely positive, and it is because they are sometimes negative in very important ways that we should ignore the role of empathy in moral decision making. But this point holds only if the negative impact of empathy cannot be modulated while retaining the positive impacts. Thus, their opposition to empathy makes sense only if empathy is considered a fixed trait.

I use the term “fixed trait” here to refer to any phenotypical trait that reliably produces identical (or near identical) effects in identical (or near identical) situations over the period of time that the trait is developmentally activated. A central feature of a trait being fixed is that it is not susceptible to change due to (nonpharmaceutical) intentional action, either by the agent who possesses the trait or by other members of the agent's society. Eye color is a standard example of a fixed trait, and gender identity is commonly thought to be a fixed trait in this sense. Major personality traits (openness, conscientiousness, extroversion, agreeableness, and neuroticism) are commonly thought to be fixed traits, and so too are such conditions as bipolar disorder and Asperger's syndrome. To be a fixed trait does not require that a trait be present throughout the organism's lifespan, and some fixed traits may emerge only developmentally as the organism matures. Prior to its emergence, the trait would be absent; after its emergence, the trait would be present and reliably activated when exposed to the requisite triggering conditions.

To say that empathy in either its cognitive or emotional forms is a fixed trait is to say that empathy is not subject to intentional behavioral modification, individually or societally. This indeed appears to be what both Bloom and Prinz are saying. Not only do both affirm versions of this thesis explicitly, as already noted, their argument against empathy holds little force if they do not subscribe to the claim that empathy is a fixed trait. But, as I will argue in the following sections, there is little reason to think this is true, and recognizing this contributes to a quite different evaluation of the role of both cognitive and emotional empathy in the moral life.

DOES SCIENCE SHOW EMPATHY IS A FIXED TRAIT?

In their critiques of empathy, both Bloom and Prinz make extensive reference to the scientific literature to buttress their claim that empathy is not only neither necessary nor sufficient for moral functioning but also frequently harmful to the goal of acting morally. In arguments concerning fixity, the first line of argument is often a genetic one, and the claim that our capacity for empathy is in part heritable is plausible. One study indicates the capacity for empathy is 68% heritable (cf. Chakrabarti and Baron‐Cohen ). Infants can imitate facial expressions within days of birth, and at 12 months babies can imitate goal‐directed actions, abilities that would require the capacity at least for cognitive empathy (Schwier et al. ). As early as age two, children engage in spontaneous other concern and comfort giving, suggesting the capacity for emotional empathy (Davidov et al. ). By 3–4 years, children begin to be able to detect false beliefs, a more sophisticated test of cognitive empathy. Indeed, the capacity for empathy is universal enough that we diagnose those who have abnormally diminished empathic capability as impaired, as in the case of autism.

From the fact that eye color is heritable, we conclude that eye color is fixed: a person born with brown eyes will have brown eyes for the rest of his or her life. Behavioral traits are typically more complex, and to say that a behavioral trait is heritable or, more precisely, partly heritable, is not to imply fixity in the same way. Notably, Bloom and Prinz do not cite genetic studies to support their position; rather, they appeal to a wide range of studies that seem to indicate ways that empathic reactions can impel us to act in a contrary to moral way. For present purposes, I will focus on a few that play a more prominent role in their (and my) argument.

Among those cited by Prinz, a couple of older studies of empathy, those conducted by C. Daniel Batson and colleagues, are among the most significant. Over a period of decades, Batson has pursued the empathy‐altruism hypothesis, the view that empathy motivates altruistic behavior (cf. Batson ). In a pioneering study, Batson and colleagues () demonstrated in an experimental condition that female undergraduate students who were primed (encouraged to believe) that they were similar to another student receiving shocks as part of a putative examination of aversive conditions were more likely to accept the opportunity to trade places with the student than those in a control condition, and the majority of those primed did this even when they had the opportunity to escape the situation. An unstated assumption of this experiment is the claim that a similarity prompt is sufficient to produce an empathetic response, and that it is the empathetic response rather than some other factor that is driving the behavior. In this case, the assumption seems reasonably plausible. An assessment of similarity would require employment of some level of cognitive empathy (thinking about the other or simulating the other), and it is a plausible though contestable hypothesis that such cognitive empathizing would induce at least weak emotional empathy. In this and several of the other studies that are discussed by Prinz and Bloom, the moral status of the putatively empathy‐induced behavior can be raised. In this case, it would clearly not be morally forbidden to offer to be shocked instead of the other student. It would also be difficult to argue that it is morally required. After all, both students signed up for the experiment voluntarily, and both could leave at any time. The shocks were unpleasant but not severe. Nevertheless, we might say it was morally good though not required to help the other student. But does it speak to good character when priming affects the likelihood of helping, as the study seems to imply? That is another question.

A much later experiment (Batson et al. ), provocatively titled “Immorality from Empathy‐Induced Altruism: When Compassion and Justice Conflict,” paints a more complex and less favorable picture of the link between empathy and altruism. The paper describes two experiments; the second is the most salient to the argument that Prinz makes, and it is the one that he cites (Prinz ). In this experiment, subjects watched a (fictitious) interview of a terminally ill child, and they were then given the opportunity to move that child up on a waiting list for a medicine that would improve the child's quality of life, though not reverse the terminal condition. Subjects were told that the professor conducting the experiment had inquired with the firm (again fictitious) administering the drug, and that the firm would allow the subject to move the child up on the waiting list, although subjects were also informed that doing so would displace children who had been on the list longer or whose condition was more serious. In this case, subjects were either given no other communication prior to the decision, a communication to be objective, or a communication to take the perspective of the child being interviewed (“imagine how the child who is interviewed feels” [Batson et al. 1995, 1048]). After listening to the interview, subjects then filled out an emotional reaction questionnaire, rating themselves on such emotions as sympathetic, warm, compassionate, or tender. When given the opportunity to help the child by moving her up the list, the majority (73%) of those who were primed with the “high empathy”/perspective‐taking communication chose to help the child by moving her up the list, while only a minority of those in the control conditions did so.

In this case, the experiment appears to measure directly the effects of priming for cognitive empathy, and such priming does appear to induce helping behavior, though the form of altruism demonstrated is clearly parochial. Further, the behavior motivated in this case does seem to be clearly immoral, violating principles of fairness in the decision to move the individual child up the list past those who had waited longer or were in more desperate need. Notably, the experiment does not directly test emotional empathy, but rather other emotions such as sympathy and compassion. One might argue that these are proxies for emotional empathy, but it is an unstated assumption, and one that is important given Prinz's endorsement of these emotions over emotional empathy.

A third study by Kogut and Ritov () and cited by Bloom concerns the relation of an identified victim effect to both empathy and moral action. In their third experimental condition described in the paper, some subjects were given the opportunity to contribute money to a single child in need of medicine or a group of eight children in such need. In each case, additional identifying information of age, name and picture were provided either singly or together. After being asked to contribute, subjects were asked to fill out a distress rating and a sympathy and compassion rating as a proxy for empathy. As Kogut and Ritov hypothesize, willingness to contribute increases as level of identification increases, and, surprisingly, willingness to contribute is larger for a single identified child than for a group of eight identified children.

Bloom argues that this experiment shows the biasing effect of empathy in a way that is immoral, but there are notable problems with this claim. First, the “empathy” self‐reports used by Kogut and Ritov are really self‐reports of sympathy and compassion. Second, these emotions seem to have no impact on the difference of giving in this case, and the results were not statistically significant. The emotional reaction associated with increased giving to an individual over the group is distress, a negative emotion that is not concerned with the other but with oneself. Second, the claim that the identifiable victim effect in this situation induces immoral behavior is debatable. Unlike the Batson 1995 study that involves principles of fairness, this study concerns beneficence, and there remains debate about obligations of beneficence and how extensive they are. In this case, we may say that the subject's giving is a good thing, while still debating whether and to what extent such giving is obligatory. But even if it is obligatory, there is a further question. Suppose I want to donate $80 and I want to maximize its impact. I could give that money to a single individual who would then receive the full $80, or I could divide it between eight children who would then receive $10 each. It is not so obvious which choice is better, and much would depend on extenuating circumstances. Unlike the Batson 1995 experiment, the moral status of the actions in the experiment is ambiguous, and it is not clear that the experiment tests for empathy in any case.

A fourth study by Xu and colleagues () and cited by Prinz focuses instead on neural responses. In this study, Caucasian U.S. subjects in the United States and ethnic Chinese subjects in China viewed videos of individuals being poked in the cheek either with a needle (painful stimulus) or Q‐tip (control). Previous studies (notably Singer et al. and ) had shown a pattern of neural activation, including the anterior cingulate cortex, associated with empathic responses, and this study showed a similar pattern, but one that was stronger when subjects were viewing same‐race individuals being poked with a needle in contrast to viewing other‐race individuals. In this study, there was no behavioral task, and there was no correlation between the strength of neural activity and responses on a self‐report empathy scale.

Does the study demonstrate a race‐based empathy response? It is a plausible interpretation that there is empathic processing at the unconscious level, but the study does not provide evidence that such race‐based empathy is present above the threshold of consciousness or that the response, whether unconscious or conscious, affects action. That being said, the study is consistent with the existing literature on implicit responses to race, and these studies do provide evidence that implicit responses, ones below the level of consciousness, can and do affect behavior (Amodio ). Those studies do not attempt to measure empathy, but it is plausible to suppose the race‐based empathetic processes or their failure play a role in the behavior seen in these studies.

The final study (Klimecki et al. ), this one mentioned by Bloom, compares the effects and neural responses of empathy and compassion training. Subjects are asked to watch a combination of low emotion and high emotion videos, the former depicting everyday scenes with low emotional content, the latter scenes of human suffering, and they were asked to do so twice, once after empathy training in which they were encouraged to resonate with the suffering of the other, and again after compassion training that aimed to cultivate “feelings of benevolence and friendliness in a state of quiet concentration.” Subjects viewed the videos while undergoing functional magnetic resonance imaging (fMRI) scanning, and they rated their emotional reaction after viewing each video. The researchers found that empathy training increased self‐reports of empathic reactions and also the strength of negative affect, while the compassion training increased self‐reports of positive affect. In this case, the reporting was consistent with the neural patterns of activation, which revealed two distinct circuits, one associated with empathic pain awareness, the other with the positive affect associated with compassion training. As interesting as this study is for mapping distinct circuits for positive and negative affect in these conditions, notably absent is the exposure of the subjects to videos of individuals experiencing positive emotions, and it is not clear how different the neural circuits would be in that case, if at all. The authors of the study note how an excess of empathy can have negative results, leading to distress and thus impeding rather than promoting behavior. The study itself, however, does not engage the subjects in a behavioral task, so it is not clear in this case whether empathy or compassion would be the more powerful motivator.

Do these studies make the case for Bloom and Prinz that empathy is on balance morally bad? Hardly. A primary problem is simply being clear whether and what kind of empathy is being examined in each case. Arguably, the experiments by Kogut and Ritov do not test for empathy at all but instead for sympathy and compassion, and the other studies test either for cognitive or emotional empathy, but never both, making comparison and the development of a systematic account difficult. Of the five studies cited, only the 1995 Batson study, conducted with a comparatively small sample of U.S. undergraduate students, provides significant evidence that an empathy prompt can induce immoral behavior. The two neuroimaging studies, as important and intriguing as they are, do not test for behavior at all. More importantly, none of these studies shows that our empathic responses are fixed, and Klimecki's study shows the opposite: our empathic responses can be tamped up with empathy training and tamped down by compassion training, among other possibilities. Rather than being fixed, our empathic reactions are responsive to interventions, intentional or otherwise.

That being said, it would be a mistake to completely dismiss the claim that our capacity for empathy and how it is extended to others is subject to various biases, whether due to nature or nurture or both. In this respect, the study by Xu et al. () is of particular interest, since its results are unfortunately quite consistent with existing research on racial and in‐group biases. It is likely that most of us learn to activate our empathic biases, both cognitive and emotional, selectively and to varying degrees, and this plausibly contributes to unjust and immoral behavior in many particular instances.

SCIENCE AND THE TRAINING OF EMPATHY

If empathy is a positive trait, to what extent can our empathic capacity be extended and improved? Over the past several years, a number of studies have begun to explore this, and here I indicate two lines of research. The first concerns the impacts of the arts on empathy. A study by David Comer Kidd and Emanuele Castano () looked at the impact of reading literary versus popular fiction on cognitive empathy (labeled in their article as theory of mind or ToM), hypothesizing that literary fiction is more demanding on the reader and thus requires greater exercise of cognitive empathy. To measure this, they used the Reading Mind in the Eyes Test (RMET), which rates the ability of an individual to accurately identify the emotional content of facial expressions, and they found that reading literary fiction did influence performance on the RMET. A similar study, conducted by Jessica Black and Jennifer L. Barnes () examined the impact of award‐winning television dramas in comparison to documentaries or viewing nothing at all, and they found that viewing the dramas had a similar impact on the RMET. A 2015 study by Loris Vezzali et al. found that reading Harry Potter novels, which treat in their own way issues of ingroups and outgroups, impacted child readers’ real‐world attitudes toward stigmatized groups.

Studies of medical students and medical residents also reveal that empathy responses are impacted by environment and possibly subject to intentional intervention and modification. In the case of medicine, the role of empathy is arguably rather complicated. On the one hand, doctors who co‐experience the pain and suffering of their patients on a regular basis would be chronically miserable, and the regular experiencing of such strong emotions can interfere with judgment and plausibly contributes to doctor burnout. On the other hand, it is important for doctors to be able to effectively employ cognitive empathy and understand what the patient is undergoing from her or his own perspective, and it is plausible that emotional empathy in optimal circumstances contributes to that process of understanding as well.

Several studies indicate lower levels of empathy in physicians and medical residents in comparison with controls. One study by Yawei Cheng et al. () shows that, unlike controls, physicians viewing videos of an individual's hands and feet being pricked by a needle do not show in fMRI scans activation of the pain matrix, implying a lack of emotionally empathic response to the viewed painful stimuli. Studies of medical students indicate that empathy declines as students progress through medical school, and one study by Mohammadreza Hojat et al. () indicates a decline in cognitive empathy in the third year in particular. Helen Riess et al. () provide evidence that training in both emotional and cognitive empathy improves empathic responses as measured by patients.

It should be emphasized that much of this work is early and ongoing. The studies of medical students and physicians imply that both cognitive and emotional empathy are subject to environmental effects and not simply fixed traits. The studies on the impacts of literature and television along with the more focused interventions of Reiss and Klimecki indicate that both cognitive and emotional empathy can be modified by intentional interventions. Exposure to literature and drama affects both empathic accuracy and the social impacts of empathy processing. Empathic accuracy is distinct from both cognitive and emotional empathy, but it is a necessary condition for both to function well. Although the RMET measures cognitive empathy, it is plausible to suppose that exposure to literature and drama engages emotional empathy: this would be consistent both with the studies of medical students and with the emotional impact of literature and drama from a first‐person perspective. Since the study by Vezzali et al. involves changed attitudes toward outgroups, it would be quite surprising if this did not include a change in emotional associations, since antipathy toward outgroups is often mediated by disgust (Buckels and Trapnell ; Nussbaum ). Further exploration is certainly needed, including the study of the duration of the effects of such interventions and their effect on character.

VIRTUES, DUAL PROCESSING, AND DYNAMIC INTEGRATIVE PROCESSING

Although Prinz argues against a positive role for empathy in the moral life, philosophers such as Michael Slote () see a positive role, and Meghan Masto () and Aaron Simmons () have to varying degrees responded to some of Prinz's arguments. Masto in particular argues that possession of empathy, including both cognitive and emotional dimensions, is an epistemic requirement for moral action, since acting in a morally appropriate way requires accurate perception of the thoughts and feelings of others, and this requires empathy. Simmons goes a bit further, arguing that empathy is sometimes sufficient for moral action, and that as such we can describe empathy as a virtue. The claim that empathy is a virtue, however, has recently been rejected by Heather Battaly (), revealing that this is territory that needs to be further explored. Like Simmons, I argue that empathy is a virtue, but it is important to keep in mind that it is just one of the virtues, an important but not sole contributor to the moral life.

Virtues are conventionally thought of as dispositions, and much has been made of the skill analogy for understanding virtue as first discussed by both Plato and Aristotle (Annas , ; Stichter ). Skills themselves are dispositions, or at least abilities, that are learned over time; being an expert carpenter or woodworker takes much time, often requiring the instruction and example of a mentor and the development of an intuitive knowledge of the craft that is not easily put into a textbook. Likewise, Aristotle argued that the virtues are themselves the result of a process of character formation impacted by moral exemplars who both teach and provide an example of the virtues that the learner is eventually to emulate. Psychologically, Aristotle's view is plausible, and while the now significant literature on expertise supports this picture of skill acquisition, studies of moral exemplars at least suggest that Aristotle is on the right track (Colby and Damon ; Gobet, Retschitski, and de Voogt ; Narvaez ; Reimer et al. ). Much of the skill literature fits comfortably into a two‐process model of mind, conceived in terms of parallel unconscious level/implicit and conscious level/explicit forms of processing, and much of this research has been focused on occasions when the two processes compete, resulting in incoherence between our actions and our stated moral judgments (e.g., Haidt ).

Although the dual‐processing model is capable of explaining a number of phenomena of interest to psychologists, it is a poor fit for an adequate understanding of the virtues. A conflicted individual, torn between the unconscious level/implicit urges of, for instance, youth and the conscious level/explicit deliberations and reasonings that one is capable of sharing with others, is hardly the model of virtue on Aristotle's account. Rather, the Aristotelian account of virtue is not one where the will overcomes a recalcitrant human nature, a view widely associated with Kant, but one where our intuitions and emotions are trained in such a way that our actions flow naturally from our dispositions rather than in spite of them. On the dual‐process model, this might simply suggest a sort of dual‐track harmony between implicit and explicit processing, but I would suggest that the real picture is more complicated, not only in Aristotle and followers like Aquinas but also in reality. Aristotle famously argues that the learning of virtue involves habit formation. In English, “habit” typically connotes the notion of “mere habit,” a routinized behavior that, once learned, becomes automatic and no longer requires thought, such as the action of reflexively looking both ways before crossing the street. While there is reason to believe that Aristotle sometimes had this in mind—he claims that we learn the virtues by doing them—his own account is richer than this, implying something we might call an active disposition, one that involves not simply an alignment of implicit and explicit mental processes, but one that dynamically integrates them as an interactive whole over time. To be courageous, then, is more than reflexively throwing oneself into the line of fire when the situation appears to require it; it involves an accurate and continually updated awareness of which situations those are, when they arise, and whether the current situation demands such action or not. It also involves recognition of when a situation requires active deliberation or not, and such capacity and awareness of the need for deliberation is the result of the honing of both our implicit and explicit processing capacities.

Aristotle famously argues for an understanding of each virtue in terms of a mean between excesses of any given character trait, and while the view can be easily caricatured, it does provide insight to understanding many virtues, including that of empathy. Virtues are not simply rules, and while the capability to follow rules is often important, rule following in the moral domain is a complex affair that involves understanding when and how a given rule applies to a given situation. Thus, being courageous involves much more than following the rule “be courageous” or “act courageously” might suggest, because the line between being courageous and cowardly, or courageous and foolhardy, is often quite fine, and it requires a similar fine‐tuning of the virtue of courage that, for most of us, requires years of experience and moral formation. But it is also the case that, for most of us, such tuning is only rarely fine and more often rough, with the result that we are sometimes too cowardly when we should have been courageous, or more rarely simply foolhardy instead.

To sum up, we can define a moral virtue as follows:

A moral virtue is an active disposition (character trait) that (1) is learned over time, (2) integrates implicit and explicit processing, and (3) is a necessary component of full moral functioning.

What makes a virtue a moral virtue is the fact that it is a necessary component of moral functioning, and this is part of what differentiates a moral virtue from a skill. If it turned out that empathy was not required for moral functioning, then it would merely be a skill and not a virtue. Skills are learned over time, and there are some skills that integrate implicit and explicit processing, including activities such as playing chess or baseball. But playing chess or playing baseball are not necessary components of the moral life, and so they are not moral virtues. As it stands, this definition may be thought too expansive, because there may be character traits that turn out to be necessary for moral functioning but do not fit our intuitions as to what should count as a virtue. Aristotle classified intellectual virtues as distinct from moral virtues; the thought may come to mind that many intellectual virtues are necessary conditions for moral functioning, but because they are intellectual virtues they should not be classed as moral virtues. But this assumes that the categories of moral and intellectual virtue are mutually exclusive. Aristotle himself noted that practical wisdom (phronesis) is deeply connected to the functioning of the moral virtues: possession of the virtues requires practical wisdom and the proper development of practical wisdom requires the virtues. Thus, even if we narrow the sphere of moral virtue, we will necessarily have to have an account of relevant nonmoral virtues that are in complex ways connected to the moral virtues. My own preference is to use the more expansive definition of moral virtue employed here, but other categorization schemes are possible.

EMPATHY AS A VIRTUE AMONG THE VIRTUES

Should we think of empathy as a virtue? Although I have been so far largely critical of Bloom's and Prinz's critique of empathy, it is important to note some areas of agreement with the model of empathy as a moral virtue being proposed here. First, Bloom and Prinz are clearly correct that the capacity and employment of empathy is not sufficient for being moral. The fact that I am both capable and willing to think about the thoughts and feelings of others does not guarantee that my behavior will be salutary, and such cognitive empathy (theory of mind/mental simulation) may be used for quite immoral purposes. Emotional empathy is not sufficient either, for although emotional empathy may well prevent me from harming others because I will also feel their harm, emotional empathy does not by itself necessarily lead to appropriate helping behavior and may indeed lead to the opposite, causing distress that one simply wants to avert. This would be true even if it were the case that our likelihood of feeling empathy was not partial or tied to ingroup/outgroup distinctions, as the studies by Batson et al. (1981, 1995) and Xu et al. () imply.

This leads to the second point: one important issue regarding empathic processing in normal subjects is its often parochial nature. The scientific evidence is consistent with the claim that, in many instances, our propensity for both cognitive and emotional empathy is limited in ways that are morally problematic, and the parochial character of our willingness to empathize is amply supported by even the most cursory survey of history. Ingroup/outgroup distinctions in particular are pervasive to human interactions, and the human capacity to dehumanize members of outgroups in ways that appear to all too easily cut off empathic processing is central to understanding the history of group violence and genocide in particular. That being said, a few cautions are in order. First, it is important to note that the issue of parochialism is an issue not of an excess of empathic processing but of its deficiency. Arguably, if human beings were not able to turn off or tune down empathy in this morally problematic way, intergroup violence and genocide would be much less likely to occur. Negative effects of the positive employment of empathy come into play only when it results in preferential treatment for a group or individual in a way contrary to what justice requires. But even here we must be careful. There is a long and ongoing debate about how we ought to balance general moral obligations and special obligations to loved ones, and only the narrowest form of act utilitarianism would demand that we not take the latter into account when calculating the greatest good for the greatest number. That empathy is sometimes parochial is not by itself a moral indictment, and part of understanding the moral role of empathy is to understand when and how such partiality goes wrong.

Third, Prinz makes a persuasive case for the claim that empathy does not apply, at least in a simple way, to all forms of moral judgment. In cases like tax‐dodging, department store theft, or environmental harms such as driving a particular plant species into extinction, there is no obvious victim to empathize with. Particularly as we move from the realm of intimate and interpersonal interaction to larger scale, group‐level interactions, including considerations of the future of the group, groups, and needs of future generations, more abstract reflection on justice, on the right and the good, is called for, and such reflection will need to draw on more than empathic processing. Put differently, every instance of moral judgment does not require online empathic processing, and so empathy is not a necessary condition for every instance of moral judgment in this narrow sense.

That empathic processing is not required for every instance of moral judgment is very different from saying that empathic processing is not a necessary condition at all, and here I disagree with Bloom and Prinz. Rather, there is good reason to believe that both cognitive and emotional empathy are globally necessary features of moral judgment and action. That this is true of cognitive empathy may seem most obvious. In order to make an accurate moral judgment and engage in the right kind of moral action, I need to have an accurate understanding of the social environment I am acting in, and this requires understanding not only what the other individuals involved are thinking and feeling but also how my actions will likely impact their future thoughts, feeling, and well‐being. I may be a teacher who has two students who have performed poorly in a paper, but one is timid and takes criticism poorly while the other is arrogant and responds only to a stronger engagement with her failings. As a good teacher, I have an obligation to engage both students in order to encourage improved performance, but the way I engage the two students will likely be very different, and it will be different based on my taking the time to consider how they will respond to my encouragements and criticisms. The need for such engagement of cognitive empathy is continual. Both Prinz and Bloom invoke the example of autism as a counterexample, but it is a perplexing one. Citing examples of high‐functioning individuals like Temple Grandin, they argue that since such individuals largely follow the norms of society and, at least sometimes, see themselves as following morality in terms of a strict code of rules, this shows that empathy is not a necessary feature of moral functioning, given that a central feature of autism and Asperger's syndrome is diminished capacity for a (cognitive) empathy. Even if Bloom and Prinz are largely correct in the details (for a critique, see Masto ), the conclusion does not follow, because it is precisely because of their impairment that such individuals have such difficulty navigating the social world, sometimes in morally relevant ways.

What then of emotional empathy? I have previously argued (Peterson ) that emotional empathy is a natural necessity of mature moral decision making and action, and we can note two ways in which this is likely true. First, there is good reason to think that emotional empathy is an important factor in the process of moral development from child to adult. Not only do we experience moral contagion from infancy on, we over time learn to interpret such contagion and what it means. Emotional empathy provides, so to speak, an important cognitive shortcut: if while playing baseball one child sees another child struck in the face with a fast pitch, the child does not need to go through the relatively ponderous process of inferring the existence of pain in the other child, the observing child simply feels the pain, more quickly absorbing the relevant lessons that follow. Emotional empathy thus plays an important role in moral learning and development over time, and this provides an important argument for the role of literature and drama in childhood education, as the experiments cited above demonstrate and which Martha Nussbaum (), among others, has recently argued.

As already noted with respect to the experiment by Klimecki et al. (), it is problematic that much of both the scientific and philosophic literature focuses only on empathy with negative states and suffering and pain. Humans are just as capable of sharing and reflecting on positive emotions as well as negative ones, and the ability to understand the positive emotions of others, to feel joy with others, to anticipate what positively motivates others, is as essential a part of the moral life as an understanding and sharing of the negative emotions of others. A person only capable of sharing the suffering of others and not their joys would be not only miserable, but likely over time barely able to function as a human being. The ability to share positive emotions is a prime feature of affiliative relationships, and much of our decision making and action necessarily incorporates into it not simply the negative but also the positive impacts of our actions. This too is a learned and developmental process: we learn to become good friends and, later, good spouses and partners and good parents. Without such awareness, our efforts fall flat or even backfire, despite the best of intentions.

Although I am emphasizing here the significance of emotional empathy for moral development, I would emphasize the same for cognitive empathy, especially since the two cannot be completely separated. Thus, even if it is the case that empathic processing is not a requirement in every instance of mental judgment, there is nevertheless a history to our moral judgments, so that while emotional and cognitive empathy may not be present in every instance, they are likely present in the history of moral development that informs how each instance is considered, with the result that both cognitive and emotional empathy are necessary conditions of moral judgment and action in this developmental sense.

In addition, emotional empathy is necessary in a stronger sense in at least some instances of moral judgment, because emotional and possibly cognitive empathy are both upstream initiators of moral judgment and action. By this, I mean that when some event occurs, the initial reaction often is or ought to be an empathetic one, whether implicit or explicit. Such a reaction then triggers a chain of psychological processes and actions. Depending on the situation, such a chain may be short or long. Figure provides one simple model of how such processing might occur. To take a paradigmatic example, if I see a child drowning in a pond, a first response likely includes an empathetic reaction, perhaps along the lines of the neural response observed by Xu et al. () and also by Cheng et al. () when subjects viewed another being poked with a needle. In such a case, the chain of causation is short, resulting in instantaneous action to save the child. But the chain may be quite a bit more complicated. When news organizations in September 2015 published the image of Aylan Kurdi, a Syrian refugee child whose body was discovered washed up on the Turkish shore, the image instantly seized global attention, prompting intense debate over the ongoing refugee crisis. It is plausible in this case to suppose that the initial reaction, at least for many, was an empathetic one. People could not empathize directly with young Kurdi, for he had already passed, but likely what many people felt was shock and horror at the sight of his body washed up on an otherwise pristine beach; as the ensuing days unfolded, no doubt many imagined what it must have been like to be that family and that child, to be adrift in the stormy Mediterranean, to lose a child after already so much heart‐rending tragedy.

Empathy in Dynamic, Integrative Processing.

But such empathetic reactions are only the beginning. Kurdi was already dead, and nothing could be done to help him, but the importance of his death was not only that he died, but that many were and (at the time of this writing) still are dying as a result of the refugee crisis stemming from the Syrian conflict. So, one's thought processes turn from Kurdi as singular individual to representative individual, and to the broader question of how one can best address the ongoing suffering spurred by the crisis. The answer to that question is quite complicated, involving current immigration policies and politics of member European Union states, projections of the causal impact of this or that policy within the EU, as well as the broader questions concerning how best to deal with the Syrian conflict, which is the proximate cause of the crisis. Such reasoning involves a knowledge of history, intimate understandings of human nature and how people and nations behave, and abstract ethical principles/rules of thumb. Although I have emphasized the initiating role of emotional and cognitive empathy, it is important to note as well the role both likely play in the ongoing process of evaluation. To understand, for instance, the implications of various immigration policy options requires an understanding of how immigrants would be received in various host countries and what life might be like for those immigrants once settled, especially given the rather mixed ability of many European countries to treat non‐European populations in a fair and equitable manner. Note here that neither emotional nor cognitive empathy necessarily leads to sympathetic evaluations in all cases: it is important to understand, for instance, the variety of negative and sometimes racist reactions to the migrant populations, but that does not imply identification or sympathy; in some cases rather the reverse.

If the preceding is correct, then it starts to become clear why it is important to think of empathy as a virtue. Evidence supports that empathy is a trait that is subject to modification through learning, and it is a trait that has clear implicit and explicit dimensions. If the above analysis is correct, it is also required for full moral functioning: most moral judgments will require the employment of cognitive empathy, and while not every moral judgment requires emotionally empathic processing, some do, and empathic processing in both forms is arguably important as well for normal moral development. If empathy is a necessary feature of moral functioning, then it is important to empathize in the right way, in the right amount, and at the right time. If empathy were not essential to moral functioning, we might say that empathy is simply a skill, useful in business negotiations among other domains. But since it is necessary for moral functioning, it makes most sense to think of empathy as a virtue, one that needs to be developed appropriately. As Bloom, Prinz, and a number of the scientific studies discussed above demonstrate, most of us have room for improvement in how and when we empathize. We may empathize too little, or may empathize too little particularly with those who are different from us, or too much with those who are like us. Our emotional empathy may be altogether too strong, as it perhaps is for those individuals unable even to watch someone else receive an injection. There are a variety of ways that we may empathize wrongly, and if we focus on cognitive empathy, we may consider ways in which we think too much or too little about others, or ways in which we have difficulty simply imagining what it is like to be the other as other. All of these limitations are features of the human condition; some of us are better are at some of these elements than others, and likely all of us are continually learning and refining and updating our capacity for empathy as we continually engage new people and populations.

Alongside of these considerations, it is important to keep in mind that empathy is only one of many virtues among the virtues, and while in much of the literature virtues are often treated separately, it is important with empathy in particular to understand it as being in relation to other virtues in a way that is also virtuous. Empathy that fails to motivate action or motivates the wrong sort of action is not much good as a virtue, and for empathy to be virtuous is for it to be connected properly to other emotions such as sympathy, compassion, and courage rather than distress and flight. It is thus understandable why in the popular imagination and, as we have seen, even in the scientific literature, empathy, compassion, and other positive emotions are often conflated, since we are not interested simply in empathy as a cognitive phenomena but as the initiator of a chain of thoughts, emotions, and actions that are connected to the moral life. Virtuous empathy does so in a positive way, but vices of empathy such as parochialism have the opposite effect.

CONCLUSION

Although I have been critical of the attacks made by Bloom and Prinz against empathy, it is important to note the positive contribution these attacks have made. A quick internet search and survey of popular titles point to a bandwagon effect associated with empathy, and like many popular psychological concepts, the result is that what is valuable about a concept becomes drowned in a sea of competing and often ungrounded claims. As Bloom and Prinz argue, empathy is not good without qualification, nor is it by itself the solution to all the world's problems. But empathy in both emotional and cognitive forms is important for moral functioning, and it is best understood as a virtue among the virtues, likely playing an essential role in moral development and the ongoing processes of moral deliberation, judgment, and action. Rather than being a trait fixed at birth, empathic capacities are shaped by the environment and subject to intentional interventions. The good news is we have evidence to support the philosophical claim that the role of empathy in the moral life can be improved; the bad news is that, as some scientific studies suggest, many of us are deficient in important ways, suggesting an important and continuing role for moral education as well as better scientific and philosophical understanding of the role that empathy plays in the moral life and the factors that shape it.

On the scientific side, more work needs to be done on the way that interventions can shape both cognitive and emotional empathy, especially over long durations, and care is needed in using consistent measures of empathy that clearly distinguish empathy in both cognitive and emotional forms from related conceptions of sympathy, which is usually construed as an emotion, and compassion, which can be construed as a cognition, an emotion, or both. On the philosophical/theological side, I have left relatively untouched questions of the relation of empathy to compassion and broader questions of the role that societies and institutions play in fostering healthy forms of empathic processing. Regarding the latter, Alasdair MacIntyre's Dependent Rational Animals () provides a starting point for thinking about the role communities play in character formation. Regarding the former, important preliminary work has been done (e.g., Burns ), and this will require not only a nuanced account of empathy, but also an understanding of compassion in all its complexity and its proper role in a larger moral framework. Just as a “stoic” cognitive empathy incapable of by itself feeling the joy of others is morally inadequate, so too a “stoic” compassion that helps without genuinely caring for the other as other falls short of the best kind of life. Empathy and compassion nurtured in communities to create individuals who are able to care across communities is the mutual task all must engage in.

ACKNOWLEDGMENTS

Versions of this paper were presented at the 2016 International Society for Science and Religion sub‐meeting of the American Academy of Religion annual meeting (November 2015) and at the South Dakota Philosophical Society meeting (November 2016). I am grateful for comments from those attending these conference sessions. I would especially like to thank Michael Spezio, Celia Deane‐Drummond, and Charlene Burns for valuable feedback. I also extend my appreciation to two anonymous reviewers who provided helpful comments requiring me to clarify several portions of the article.