21.1.1 Criminal Blame and the Myth of Free Choice

Enlightenment and post-Enlightenment legal theorists did not initially focus on the social context of criminal law. Philosophy and economics dominated earlier theorizing about law and punishment, and a focus was on the legitimacy of permitting the state to use force against the citizenry, depriving individuals of liberty in the process. Deprivation of liberty was therefore reserved for seriously harmful conduct deemed worthy of societal moral condemnation; conduct that does not rise to the level of societal moral condemnation is dealt with by the civil side of the legal system, outside of criminal law. This definition separates crime from other conduct deemed too trivial to rise to the level of a criminal offense (for example, negligently breaking a window – possibly a civil offense but not a criminal one – or rudely slamming the door in someone’s face).

The state’s power to deprive liberty and attach stigma is justified traditionally by a constellation of principles referred to by the term of art “legality.” Under these principles, the state is justified in imposing criminal liability and punishment when it gives fair notice as to which conduct is prohibited. These conditions enable each individual to make a choice at any given point in time about whether to comply with the law or not (Duff, Reference Duff1993). So long as the rules are made public, are sufficiently specific, are announced in advance, and are promulgated preferably by a democratically elected branch such as a legislature, each person has sufficient opportunity to freely choose to comply or to face the consequences of not complying. The background assumptions of this model center on rational individuals exercising their free will after carefully weighing the risks and benefits of engaging in unlawful conduct. There are, to be sure, various ways in which this Enlightenment notion of rational actors making free choices is largely fictional. Law governs a complex array of human activity, and law does not generally influence individual behavior in a vacuum, as earlier philosophers generally assumed. Instead, group identity, social norms, and various social motivations interact with law to provide motivations to comply (Nadler, Reference Nadler2017).

The view of law as a coercive tool to shape behavior implied that the function of legal blame is to demarcate instances in which individuals had engaged in prohibited conduct. In the domain of criminal punishment, much emphasis was placed on the extent and circumstances under which punishment effectively deters, and in many domains, deterrence indeed plays a critical role in keeping undesirable behavior in check. For example, we rely on the threat of legal sanctions to prevent offenses like theft, even though moral values contribute as well. Most people would desist from cheating or stealing where doing so would clearly and directly harm another; at the same time many people are tempted and at times do cheat or steal when they are able to tell themselves a story that erases the victim or the harm (Feldman, Reference Feldman2018; Nadler, Reference Nadler2020). Specifying the circumstances under which the threat of criminal punishment is an effective deterrent is an important area of law and economics, and while there is still much yet to understand, there is a large body of well-established knowledge in this field (Becker, Reference Becker1968; Posner, Reference Posner1985; Shavell, Reference Shavell1985).

At the same time, criminal law functions through means other than threatened sanctions. When the criminal law is perceived as legitimate, it produces a sense of obligation to obey (Papachristos et al., Reference Papachristos, Meares and Fagan2012; Tyler & Jackson, Reference Tyler and Jackson2014); conversely, when particular decisions or laws are perceived as unjust, compliance in everyday life can decrease as a result (Mullen & Nadler, Reference Mullen and Nadler2008; Nadler, Reference Nadler2005). For example, people – including violent offenders – who view police and prosecutors as being honest and even-handed are less likely to commit crime in the future (Papachristos et al., Reference Papachristos, Meares and Fagan2012). And people who learn of an unjust legal result (such as an unjust jury verdict) are more likely to flout the law in their everyday life (such as by stealing a small item when no one is looking) (Mullen & Nadler, Reference Mullen and Nadler2008; Nadler, Reference Nadler2005). Qualities of the law itself, including the extent to which it is perceived as furthering justice or reflecting community values, influence the extent to which people feel generally bound by law. One ambition of criminal law in particular is to reflect the shared moral culture that forms the basis for social coordination and solidarity (Durkheim, 1893/Reference Durkheim1964; Kleinfeld, Reference Kleinfeld2015).Footnote ¹ By examining psychological processes of blame for wrongdoing, we might gain insight into the development of the shared moral culture upon which criminal law is based. Before doing so, we review the basic structure of criminal blame.

21.1.2 Actus Reus, Legal Attempt, and Blame

In Anglo-American legal systems, criminal blame depends on a few distinct components: actus reus, or guilty act; mens rea, or guilty mind; and for offenses where there is a prohibited outcome (such as death in homicide offenses), a causal connection between the act and the outcome. A basic tenet of criminal law theory is that we do not impose blame solely on the basis of bad thoughts in the absence of conduct. For example, suppose Jane conceives a plan to rob a bank. She tells a friend about the plan but does not take any action to carry it out. Jane is not criminally blameworthy because she took no action. The criminal law declines to punish solely for bad thoughts for a few reasons. First, bad thoughts are ubiquitous, in the sense that each of us has bad thoughts sometimes. Even assuming we could accurately detect instances of thinking about wrongdoing, imposing blame according to such a low threshold would brand most, if not all, members of society criminal. The second reason that criminal law generallyFootnote ² refrains from blame in the absence of conduct is that it is difficult for the state to accurately detect and assess bad thoughts. Even when a person communicates their thinking about wrongdoing in a way that reveals it to others (through, for example, a diary or a conversation), it can be difficult to know whether the actor was serious or just blowing off steam or joking or fantasizing. Finally, even if the actor is serious about engaging in prohibited conduct at the time she has the bad thought, the law recognizes that sometimes bad thoughts are fleeting, and we come to our senses before acting. The criminal law encourages actors to move past bad thoughts by not blaming until and unless the actor engages in some conduct toward the commission of the wrongful act. Imposing blame prior to conduct disincentivizes rethinking and desisting.

Criminal blame is imposed when a person goes beyond the stage of merely thinking a bad thought. One policy question that criminal law must resolve is how far beyond thought and toward completed conduct must a person go to be criminally liable for attempting to break the law? The draw of criminal attempt imposes blame when a person intends and initiates an offense but does not successfully complete it. Criminal attempt belongs to a group of doctrines (along with conspiracy, solicitation, and the like) that imposes liability before there is necessarily any concrete harm. The theory is that these acts are blameworthy because they impose social harm. We do not want to oblige law enforcement to wait until there is concrete harm before they can intervene to stop a wrongful act, and we want to be able to deal with dangerous persons before they do real damage. We also do not want to reward moral luck – for example, an assassin who fires but misses his target has done no physical harm but should not be absolved based on chance factors beyond his control.

There are two approaches to pinpointing the juncture beyond which a person’s thoughts and preliminary actions constitute a criminal offense. The more traditional approach is the “proximity” test. The proximity test asks whether the person’s conduct constitutes “mere preparation” – which is not considered sufficient to constitute criminal conduct – or whether instead the conduct came sufficiently close to the completed offense to be blameworthy. Note that the proximity test focuses on how close the person came to committing a completed crime. Under the proximity test, a would-be arsonist who makes a mental plan to burn a building and is caught with a can of gasoline and combustible materials in his garage might not be criminally liable because these acts were not in close proximity to the building and the fire – rather this conduct is arguably mere preparation. But if the person is caught as he is lighting the match while standing over poured gasoline at the site of the intended arson he would be liable for attempted arson.

The second approach to deciding how much conduct is sufficient for an attempt crime examines the question from the opposite side and asks not how close was the actor’s conduct to the completed act, but rather how far the actor went beyond mere thought. This approach is called the substantial step test and asks whether the actor’s conduct constitutes a substantial step in a course of conduct planned to culminate in a crime. Once a person has formed an intent to commit the crime and has engaged in substantial conduct toward that end, they are liable for attempting to commit the crime they intended. Under the substantial step test, liability for attempted criminal conduct attaches at an earlier time, and conduct that would not give rise to liability under the more traditional proximity test could give rise to liability under the substantial step test. In the arson example, the person caught with a can of gasoline and combustible materials in his garage might well be liable for the crime of attempted arson under the substantial step test (but not under the proximity test). One advantage of the substantial step test is that it permits law enforcement to intervene at an earlier stage when the proximity to danger is more attenuated. One disadvantage is that there is a greater chance that an innocent actor’s conduct will be misconstrued. For both the proximity test and the substantial step test, the possibility of blaming an innocent person for attempt is shielded by a separate requirement of proof of intent to commit the criminal offense in question.

There has been some empirical investigation of intuitive blame with respect to incomplete wrongdoing, both outside and within the legal context. Outside of the legal context, observers do blame an individual who is merely thinking about engaging in a harmful act, but they blame more when the individual intends to carry out the act, and they blame the most when the act is in fact carried out (Guglielmo & Malle, Reference Guglielmo and Malle2019). Within the legal context, intuitive blame seems to track the proximity doctrine of criminal liability for attempt, more than the substantial step doctrine. One study presented participants with a story about a locksmith who decides to steal rare coins from a safe in a shop. Participants chose the point of conduct in the storyline where blame should attach (locksmith enters the shop to find the safe; locksmith tries to crack the safe; and so forth) (Robinson & Darley, Reference Robinson and Darley1995). Most participants assigned blame only after the locksmith reached the point of dangerous proximity to harm – when he began cracking the safe. A vast majority thought that “casing” the shop to look for the safe is not sufficient for blame and punishment. This finding aligns more with the more traditional proximity rule for assigning blame for criminal attempt, rather than the more modern substantial step test.

A follow-up study testing various robbery and murder cases yielded similar results: laypeople tend to wait until the wrongdoer is at the point of dangerous proximity before they attach blame (Darley et al., Reference Darley, Sanderson and LaMantia1996). The studies just discussed examined the lay notion of blame for attempt, but a related question is how psychological processes of decision makers interact with existing legal standards. Mock juror studies that manipulated the legal standard for attempt (“engaged in a substantial step … strongly corroborative of intent” versus “conduct that came dangerously close or very near to committing…”) found that in ambiguous cases the proximity test resulted in a greater likelihood of blaming than the substantial step test (Sood, Reference Sood2019, Study 1). And a study that examined the role of anti-Muslim bias in judging attempt liability found that mock jurors blamed a Muslim defendant much more harshly (compared with Christian or control) when deciding attempt liability under the proximity test, but not under the substantial step test (Sood, Reference Sood2019, Study 3a). For the Muslim defendant only, the proximity test language gave rise to negative inferences not only about ultimate judgments of legal guilt but also about intent to commit the crime. Mere exposure to the language of the proximity test caused participants who judged the Muslim defendant (compared to those who judged the Christian defendant) to indicate more hostile attitudes toward Islam; no such difference emerged for participants exposed to the language of the substantial step test. These findings suggest that the stereotypical nature of the crime of terrorism combined with the Muslim identity of the defendant were primed by the language of the proximity test (“dangerously close”) but not by the substantial step test.Footnote ³ In addition to sufficiency of conduct, a critical question for questions of criminal blame is mental state, which we turn to next.

21.1.3 Mental States and Blame

21.2.1 Moral Character as Information about Mental State

At first glance, the criminal law’s risk-awareness framework does not seem to comport particularly well with the way laypeople blame intuitively in situations of unintended harm. When we judge unintended harm in everyday life, we might be somewhat concerned with whether the actor was aware of a specifically enumerated risk (in the earlier example, risk of causing death), but we are often concerned with the reason the actor decided to engage in the risky behavior and, even further, what kind of person the actor is. In general, when we judge bad outcomes in everyday life, we care about the actor’s motive and their character (Siegel et al., Reference Siegel, Crockett and Dolan2017; Uhlmann et al., Reference Uhlmann, Pizarro and Diermeier2015), in addition to judging the appropriateness of the conduct and awareness of the specific risk of harm that ended up occurring. Character-based theories of moral blame hold that the motivation to evaluate others’ character is a fundamental feature of human social cognition (Pizarro & Tannenbaum, Reference Pizarro, Tannenbaum, Mikulincer and Shaver2012). For example, Alicke’s culpable control model assumes that we constantly evaluate other people to determine which ones are trustworthy, that is, will promote rather than threaten our physical and psychological well-being (Abele & Wojciszke, Reference Abele, Wojciszke, Reis and Judd2014; Alicke, Reference Alicke2014). According to character-based theories of moral blame, we spontaneously evaluate wrongdoing based on features of character before having the opportunity to carefully consider the legally central features of mental state such as conscious disregard of risk. It is important to note that some theorists dispute that early spontaneous evaluations of character influence blame independently of diagnostic inferences about mental state, conduct, and causality (Malle et al., Reference Malle, Guglielmo and Monroe2014).

Experimental work suggests that blame is indeed susceptible to evaluations of the actor’s moral character. In one study, a person named Sam illicitly stored flammable oxygen that led to an accidental explosion and death of a bystander. Observers blamed Sam more severely if he stored the oxygen to cheat in athletics than if he stored it to start a business or to care for his sick daughter (Nadler & McDonnell, Reference Nadler and McDonnell2012). This is despite the fact that the legally relevant mens rea did not vary. Sam’s awareness that storing oxygen could be dangerous to human life would form the basis for recklessness, and there were no detectable differences in foreseeability of harm between Sam the cheater and Sam the caregiver. Similarly, in another study, participants blamed Frank more severely for causing a deadly fire when he was storing chemicals for a meth lab compared to when he was storing the same chemicals for a flower greenhouse (Nadler & McDonnell, Reference Nadler and McDonnell2012). And after John caused a traffic accident, he was blamed more if he was rushing home to hide drugs than if he was rushing home to hide a present (Alicke, Reference Alicke1992). Thus, we blame more severely for resulting harm when an actor undertakes risky conduct for questionable reasons than for laudable reasons.

Interestingly, the law of criminal recklessness indeed reflects intuitive blame’s sensitivity to reasons for acting. The most influential source for contemporary American criminal code definitions is the Model Penal Code, which defines recklessness as the disregard of a substantial and unjustifiable risk of causing a prohibited result. The substantiality requirement exists to exclude criminal blame in cases of risks that are remote or unlikely. The justifiability exemption exists to permit conduct, such as a risky surgery designed to save life. Both terms – substantial and unjustifiable – raise questions about which risks are substantial and which are unjustifiable. To guide decision makers, the Model Penal Code definition of recklessness includes the following explanatory language: “the risk must be of such a nature and degree that, considering the nature and purpose of the actor’s conduct and the circumstances known to him, its disregard involves a gross deviation from the standard of conduct that a law-abiding person would observe in the actor’s situation.”

This language helps make sense of some of the experimental results just discussed. Frank was blamed more severely for causing a deadly fire when he was storing chemicals for a meth lab than for a greenhouse. Was the risk posed by storing the chemicals substantial and unjustifiable? To answer that question, we consider the nature and purpose of Frank’s conduct. In the meth lab story, the purpose of the conduct of storing chemicals was to engage in later conduct that is both dangerous and criminal. Frank’s disregard of the risk therefore involves a gross deviation from the standard of conduct that a law-abiding person would observe. By contrast, Frank the gardener’s decision to store chemicals has a nature and purpose that is more benign and is arguably consistent with the standard of conduct that a law-abiding person would observe. The difference in blame judgments that emerged in the experiment is accounted for by the legal definition of recklessness. The same parallel between legal and intuitive blame can be seen in Sam’s decision to store oxygen for cheating or for his sick daughter, and John’s speeding home to hide drugs or a present. In sum, the legal definition of criminal recklessness is constructed in a way that reflects the tendency to intuitively blame according to the nature and purpose of the conduct that led to the harmful result.

The nature and purpose of conduct standard from criminal law recklessness arguably is more concerned with risk regulation than it is with judging moral character. A fundamental truth of contemporary social life is that we agree to live with a variety of risks, and some of these are nonnegligible. Riding in a vehicle, walking, or biking on a public way are illustrative – traffic deaths in the United States alone number over 30,000 every year. We collectively accept this risk in exchange for the social benefit derived. Evaluating criminal recklessness requires consideration of the nature and purpose of the actor’s conduct to decide whether the risk consciously disregarded by the actor provided sufficient social benefit as to exempt the actor from this particular level of blame. The recklessness judgment of blame, arguably, is not of the actor’s character but the circumstances of risk taking. Under this interpretation, Sam, Frank, and John in the studies referenced earlier were blamed more severely not necessarily because their moral character was lacking but more because their risky conduct was insufficiently justified by socially beneficial goals (Guglielmo, Reference Guglielmo2015).

It is possible, then, that in the studies discussed earlier the spontaneous evaluation of wrongdoing is not focused solely on primarily moral character, because the specific factor that varied was not moral character (or at least not solely moral character) but rather the actor’s motive for engaging in the hazardous conduct. But in other studies, the circumstances of the risky conduct indicative of motive were held constant while the moral character of the actor was varied, and the predicted relationship between bad character and more severe blame again emerged. For example, Sara’s unruly dogs escaped her yard while she slept and mauled a child to death. Sara was blamed more severely for the death if she ignored her family, ate junk food, and smoked than if she was fit, healthy, and volunteered for charities (Nadler & McDonnell, Reference Nadler and McDonnell2012). There was no effect observed of moral character on perceptions of foreseeability of death, so the relevant conscious awareness of risk was apparently unaffected by moral character, and thus there are no apparent differences in the mens rea of recklessness that could explain differences in blame judgments apart from moral character. Similarly, Nathan was a young man skiing out of control for fun who collided with another skier, causing their death. Nathan was blamed more for the death if he was an unreliable worker and rarely helped his family business than if he was a model employee and helped his family business (Nadler, Reference Nadler2012). These vignettes arguably provide stronger, more direct evidence for the hypothesis that intuitive blame is sometimes character based (Pizarro & Tannenbaum, Reference Pizarro, Tannenbaum, Mikulincer and Shaver2012).

According to a character-based account of blame, we consider character information partly because we rarely have precise information about the actor’s mental state, especially in cases of disregard of risk. Knowing something about the actor’s previous behavior – reflected in traits like fairness, kindness, trustworthiness, and integrity – help us assess what they were doing and why at the time the harm occurred. The character-based account of blame helps make sense of more severe blame in certain cases of lesser direct harm. In one study, a company manager cut vacation days only of the 20 percent of employees who are African American because he is bigoted, or 100 percent of employees (20 percent of whom are African American) because he is misanthropic. Compared to the misanthrope, the bigoted manager who made cuts based on employee race was blamed more, his conduct was perceived as more diagnostic of his character, and diagnosticity judgments were correlated with blameworthiness judgments (Pizarro & Tannenbaum, Reference Pizarro, Tannenbaum, Mikulincer and Shaver2012).

If determining moral character of others is a fundamental human motivation, then blaming might be as much about the question of “is this person good or bad?” as it is about “is this act right or wrong?” (Uhlmann et al., Reference Uhlmann, Pizarro and Diermeier2015). To the extent that these two questions are two components of blame, then moral evaluations of act and character might sometimes diverge. Supporting this notion, one study found that although animal cruelty is viewed as less immoral than similar violence toward humans, the former can be more indicative of perceived moral character. Participants read about a man who found out his girlfriend had been unfaithful and either reacted violently toward his girlfriend or reacted violently toward his girlfriend’s cat. Participants judged the cat beater’s actions to be less wrong than the woman beater’s actions, but the cat beater was perceived as having worse moral character than the woman beater (Tannenbaum et al., Reference Tannenbaum, Uhlmann and Diermeier2011), suggesting that the badness of the act and the badness of the person are separate judgments in the process of blame.

21.2.2 Intuitive Blame, Character, and Mental State

To what extent do processes of intuitive blame inform and influence formal legal processes of blame? If there is a fundamental human motivation to punish those with bad character, then perceptions of character might influence blame directly, as suggested earlier; in addition, perceptions of character might influence judgments of important elements of legal proof such as an actor’s mental state and causation, which in turn influence blame judgments (Chapter 15, this volume). Recall John, the driver who caused a car accident when rushing home. Participants perceived John to be not only more blameworthy but also more of the cause of the accident when he was going to hide cocaine than when he was going to hide a present (Alicke, Reference Alicke1992). And participants judged Nathan to be not only more blameworthy but also to have acted more intentionally in killing the other skier when he was an unreliable employee and unhelpful son than when he was the opposite (Nadler, Reference Nadler2012). Participants judged Sara to not only be more blameworthy but also to have acted more intentionally toward her dogs’ mauling of the child when she was an asocial, unfit eater of junk food than when she was the opposite (Nadler & McDonnell, Reference Nadler and McDonnell2012). The actor’s conduct was identical, and the physical chain of events that led to the harm was identical; the only thing that varied was the actor’s character or reason for acting. In these studies, bad character led perceivers to infer more intentionality and more causality than good character.

This brings us back to criminal law. The law imposes more severe blame when mens rea is more culpable, all things being equal. Some of the studies discussed earlier suggest that the moral character of the actor, apart from that actor’s motive or reason for acting, plays an important role in inferences about mens rea (such as the awareness of risk required for recklessness) and overall blame. Compared to a virtuous person, we blame a morally flawed person more harshly and we bolster these harsh blame judgments with increased perceptions of the actor’s causal role and intent to cause harm. Because it is often difficult to glean another person’s mental state with the precision that criminal law demands, the process of inferring an actor’s mental state under conditions of uncertainty might be prone to the influence of character information.

The implications for legal decision making are notable. Recall that the threshold requirement for criminal recklessness is conscious disregard of a specific risk – not risk of harm in general but rather the risk of the result prohibited by the offense. For example, if we think that a distracted driver might have been aware of the risk of causing death, but we are not sure, then learning that the driver is a person of poor moral character might be enough to push us toward inferring awareness of the prohibited risk, resulting in a more severe blame judgment. Conversely, knowing the driver is an otherwise virtuous person might pull us in the other direction, toward less severe blame. In this way, moral character might serve as a kind of proxy for mental state, so that a person with a bad character is blamed as if he were reckless, and a person with a good character is blamed as if he were not reckless. Experimental results from two studies support this possibility. In the study discussed earlier involving Nathan the skier, Nathan’s awareness of risk was manipulated as one independent variable (aware or unaware), and his character traits were manipulated as another (unreliable employee, unhelpful son, or the opposite) (Nadler, Reference Nadler2012). Bad Nathan, who was unaware of the risk (and so, as a matter of law not reckless), was blamed at about the same level of severity as Good Nathan, who was aware of the risk. That is, having bad character traits served as a kind of substitute for reckless mental state, even when the actor was explicitly described as being unaware of the risk. The same pattern of results emerged for Bad Sara, whose dogs mauled and killed a child (Nadler & McDonnell, Reference Nadler and McDonnell2012). At the same time, it is worth noting that these results are also consistent with the inverse hypothesis that good character mitigates blame. In the end, it is likely that both processes are possible, with good character sometimes mitigating blame and bad character sometimes exacerbating blame. This possibility is suggested by the data in the earlier vignette about Sam, who stored oxygen that exploded (Nadler & McDonnell, Reference Nadler and McDonnell2012). This experiment included a control group in which Sam’s motive (and thus inferences about character) was described as neutral. Compared to Neutral Sam, participants blamed Bad Sam more and Good Sam less. Of course, this experiment manipulated motive (reason for storing oxygen) rather than character directly, so more investigation is needed to explore the mitigating and exacerbating influence of character on blame.

The law of recklessness, on the other hand, assumes that decision makers will make a threshold judgment about whether the actor was aware of the risk of causing the prohibited result. Only after the actor was determined to have consciously disregarded the risk of prohibited harm can decision makers move to the next questions of deciding whether the risk was substantial and unjustifiable. During the substantiality and unjustifiability determination, observers judging legal blame consider the nature and purpose of the conduct and whether the actor deviated from that of a law-abiding person. But this all assumes an initial determination that the actor was aware of the risk. In the real world, awareness of risk, like other mental states, is rarely clear to observers. The studies discussed suggest that moral character, as well as reasons for acting, sometimes inform the threshold judgment of risk consciousness, contrary to the requirements of the legal standard.

At the same time, the law does anticipate decision makers’ tendency to consider and even overweight information about character and propensity for evil. As discussed in greater detail in Section 21.3, the law of evidence prevents the jury from learning about a criminal defendant’s prior crimes or bad acts when those previous acts are presented to show that the defendant has a propensity to engage in wrongdoing. Empirical evidence supports the concern that the jury will use prior crimes as an additional reason to blame the defendant for the current accusation (Eisenberg & Hans, Reference Eisenberg and Hans2009; Greene & Dodge, Reference Greene and Dodge1995; Lloyd-Bostock, Reference Lloyd-Bostock2000; Wissler & Saks, Reference Wissler and Saks1985). But the findings on moral character’s influence on blame reach far beyond the traditional concerns with propensity evidence that excludes information about bad acts or similar crimes. The influence of prior crime and acts on legal blame diminishes substantially when the prior crime was minor or dissimilar to the current offense. But the moral character studies discussed earlier show that when we size someone up as a bad person – even through relatively subtle cues like lack of generosity and unreliability – we perceive their unintentional acts as more causal, their mental states more intentional, and their blameworthiness greater than a similarly situated good person. The subtlety of the manipulated traits (irresponsible worker, smoking, eating junk food) demonstrates that we perceive badness not only in people who have engaged in serious past wrongdoing but also in people whose common, everyday conduct indicates a lack of concern for group well-being.

Although the character-focused nature of blame for unintended harms is in tension with contemporary definitions of recklessness that follow the Model Penal Code, such character-based blaming processes are more consistent with common law notions of mens rea that fell out of favor over the course of the twentieth century. Before the Model Penal Code prompted many American state legislatures to update their criminal codes, criminal offenses were composed holistically rather than as a collection of component parts. English common law used terms like “malicious,” “wanton,” or “depraved” when referring to mens rea. The language encouraged character assessments, and courts interpreted mens rea as indicative not only of evil intent but evil character (Pillsbury, Reference Pillsbury2000). For example, a judge writing in the late 1800s – in seeking to distinguish evil passions from legitimate excuses – declared that evil passions “are the outpourings of a wicked nature, not of an unsound or disabled mind” (Pillsbury, Reference Pillsbury2000, p. 84). Until the modernization brought about by the Model Penal Code and associated theorizing in criminal law and criminology, the standard for an unintentional homicide that qualified for elevation from manslaughter to murder was defined as killing that demonstrates “wickedness of disposition, hardness of heart, cruelty, recklessness of consequences, and a mind regardless of social duty” (Pillsbury, Reference Pillsbury2000 p. 162). As with intuitive blame, the focus under common law crime standards was on moral traits and character more centrally than state of mind.

Does the contemporary cognitive turn in criminal law away from a central focus on character and toward assessment of the actor’s cognition matter? Arguably in some cases the focus of the standard – character versus cognition – makes a difference. Consider, for example, the real-life case of Marjorie Knoller and Robert Noel, who kept two dogs, weighing 150 and 130 pounds respectively, in their San Francisco apartment. The dogs repeatedly bit neighbors and attacked other dogs, and Knoller admitted that she did not have the strength to control them. The dogs escaped and mauled to death a neighbor down the hall as she was entering her own apartment. The jury found Knoller guilty of murder because she caused a death while acting with “an abandoned and malignant heart” – the old character-based standard. At the time, the trial judge clarified that this standard required proof that Knoller acted with an awareness of endangering human life – the more contemporary cognitive standard. But in an unusual move, the trial judge set aside the jury’s verdict of guilty on the grounds that Knoller was not in fact aware of endangering a human life because Knoller – a lawyer herself – claimed that she did not anticipate that her “gentle and loving and affectionate” dogs would ever kill someone. The judge decided that no reasonable jury could have found beyond a reasonable doubt that Knoller was aware of the risk that her dogs would cause death. The trial judge used the contemporary cognitive focus – awareness of risk of death – to exonerate Knoller.

The distinction between the older character-based standard of acting with an abandoned and malignant heart on the one hand, and the contemporary cognitive standard of awareness of a substantial risk of death on the other, becomes even more stark when considering what we know – and what the jury also knew – about Knoller’s moral character. The veterinarian who examined the dogs at the time Knoller adopted them wrote her a letter warning her that they were liable to attack and maim. The handler who fostered the dogs prior to turning them over to Knoller conveyed a similar warning. During each of many incidents of aggressive behavior by the dogs toward neighbors, Knoller and Noel not only refused to apologize but treated the victims with disdain and hostility. For example, after one neighbor complained that one of the dogs bit his rear end, Noel replied, “um, interesting.” There was some suggestion that the couple acquired the dogs as part of a breeding operation for fighting dogs – their correspondence with their business partners described one of the dogs as “Wardog” and “Bringer of Death: Ruin: Destruction.” Also, Knoller’s and Noel’s business partners were inmates in a state penitentiary and described by the court as members of the Aryan Brotherhood prison gang.

Although the trial judge had vacated the reckless murder conviction based on the cognitive awareness of risk standard, the case did not end there. After an appellate court reversed, a new court reinstated Knoller’s conviction for reckless murder, which comported with the jury’s initial decision, and also arguably with a character-based assessment of blame. A character-based blame assessment would focus on the older common law standard – whether she acted with an abandoned and malignant heart. Considering Knoller’s callous, contemptuous behavior toward members of her community, as well as her partnership with white supremacists apparently in furtherance of breeding fighting dogs, blame for the elevated offense of reckless murder (as opposed to a lesser offense of manslaughter) appears easier to justify.

When a person who caused harm is charged with a criminal offense, the severity of blame depends on inferences about intentionality, awareness of risk, and reasonableness, among other things, and prosecutors, judges, and jurors are free under character-based standards for blame to infer wicked subjective culpability when judging members of minority racial, ethnic, and religious groups. Empirical work on inferring mental states of wrongdoers suggests that the race of the actor can influence these inferences. For example, observers who watched two people in a heated discussion followed by one person ambiguously “bumping” his body into the other were more likely to describe the bump as a “violent shove” and to make dispositional attributions if the actor was Black rather than White (Duncan, Reference Duncan1976). When the protagonist was White, observers were more likely to excuse the conduct as “horsing around” and to make situational attributions to explain it. Further, in the criminal justice context, Black boys are perceived as older, less innocent, more agentic, and more responsible for their actions than White boys (Goff et al., Reference Goff, Jackson, Di Leone, Culotta and DiTomasso2014). In this sense, race itself can stand in for character: “In cases involving Black defendants, their pigmentation and identity performance are proof of their bad character or criminal propensity” (Armour, Reference Armour2020, p. 166). If true, the older common law mental state terms might serve to exacerbate racial and other biases. More work is needed to explore the extent to which more general common law terms are susceptible to biased decision making. For example, does asking jurors whether a Black defendant or a Muslim defendant had a “vicious will” or “abandoned and malignant heart” at the time he caused death provide a greater role for biased decision making than asking them whether he consciously disregarded a risk or acted with extreme indifference to human life? These are empirical questions yet to be explored.

22.1.1 Moral Foundations Theory: Morality Causes Political Views

Moral foundations theory (MFT; Graham et al., Reference Graham, Haidt, Koleva, Motyl, Iyer, Wojcik and Ditto2013), today’s predominant account of the relationship between moral and political views, posits that biological, cultural, and developmental differences determine what people count as morally relevant, and in turn shape their political identities.

According to MFT, people evolved modular intuitions, such that they can feel moral concern in five different domains: 1) care/harm, 2) fairness/cheating, 3) loyalty/betrayal, 4) authority/subversion, and 5) purity/degradation (liberty/oppression may be a sixth domain; Graham et al., Reference Graham, Haidt, Koleva, Motyl, Iyer, Wojcik and Ditto2013). These intuitions are activated to different degrees in different people, in part depending on biological predispositions to feel specific moral emotions more intensely (Inbar et al., Reference Inbar, Pizarro, Iyer and Haidt2012); for instance, disgust is specifically tied to judgments of purity (Tracy et al., Reference Tracy, Steckler and Heltzel2019). These intuitions are also activated differently depending on things like sensitivity to threats and sociocultural factors (Graham et al., Reference Graham, Haidt and Nosek2009, Reference Graham, Haidt, Koleva, Motyl, Iyer, Wojcik and Ditto2013). Thus, a child biologically more sensitive to disgust is more likely to moralize purity; one raised by vegans who taught compassion for animals is more likely to moralize care.

By the MFT account, people adopt political beliefs that appeal to their most activated moral foundations. In some people, the care and fairness foundations (called the individualizing foundations, because they concern individual rights and freedoms) are much more strongly activated than the others. These people are drawn to liberal policies like protecting the welfare of all people, regardless of their identity. In others, the purity, loyalty, and authority foundations (called the binding foundations because they unite people into larger, cohesive groups) are activated almost as strongly as the individualizing ones. These people are drawn to conservative policies that prioritize the strength and welfare of their in-group. Returning to the earlier example, the disgust-sensitive child likely has more conservative attitudes about sexuality (Inbar et al., Reference Inbar, Pizarro, Iyer and Haidt2012), while the vegans’ child likely has attitudes about factory farming more characteristic of liberals.

Supporting this account, liberals and conservatives reliably differ in their endorsement of the binding foundations. Across hundreds of thousands of people from around the world (Graham et al., Reference Graham, Haidt and Nosek2009, Reference Graham, Nosek, Haidt, Iyer, Koleva and Ditto2011), conservatives ascribed greater moral relevance than liberals did to considerations such as “whether or not someone did something to betray his or her group,” “whether or not someone showed a lack of respect for authority,” and “whether or not someone did something disgusting,” compared to liberals. Likewise, conservatives demanded more money to consider behaving in disloyal, impure, and subversive ways (Graham et al., Reference Graham, Haidt and Nosek2009); for example, they demanded $10,000 to blaspheme their parents, while liberals would do it for only $600. And conservatives reference loyalty, authority, and purity more when describing peak experiences and turning points in their lives (McAdams et al., Reference McAdams, Albaugh, Farber, Daniels, Logan and Olson2008). In these studies, liberals also endorse the individualizing foundations somewhat more than conservatives, though these differences are generally smaller.

These robust differences help explain liberals’ and conservatives’ diverging policy attitudes. Across two studies totaling nearly 25,000 participants (Koleva et al., Reference Koleva, Graham, Iyer, Ditto and Haidt2012), people’s moral foundation profile predicted their stance on all political issues surveyed. For example, the more people endorsed the care foundation, the more they opposed the death penalty; likewise, those endorsing the purity foundation disapproved more of impure sexual acts (e.g., casual sex, pornography), same-sex relationships, and impure genetic practices (e.g., cloning).

Despite the mountain of research MFT has generated, its detractors note that MFT’s most popular measure conflates the moral foundations with well-known liberal and conservative differences (e.g., pride for one’s country’s history; Kugler et al., Reference Kugler, Jost and Noorbaloochi2014) and thus may have misidentified the true core differences in morality that cause political disagreements. Moreover, MFT’s proponents have rarely measured experimental effects of morality on political views, thus leaving open the possibility that they have misidentified the direction of causality (Bakker et al., Reference Bakker, Lelkes and Malka2021; Hatemi et al., Reference Hatemi, Crabtree and Smith2019). In light of these criticisms, we consider two alternative accounts that suggest morality affects political ideology, as well as a third set of ideas that suggests political views may shape morality, rather than the reverse.

22.1.2 Two Alternative Accounts of How Morality Fuels Political Beliefs

Among the theories that challenge MFT’s account of moral differences without disputing that these likely cause political disagreements, two notable ones are the model of moral motives (MMM; Janoff-Bulman & Carnes, Reference Janoff-Bulman and Carnes2013) and the affective harm account (AHA; Gray et al., Reference Gray, MacCormack, Henry, Banks, Schein, Armstrong-Carter, Abrams and Muscatell2022). While these models differ in important ways, both propose novel frameworks to understand human moral concerns and, in doing so, argue against MFT’s view that liberals and conservatives are morally mismatched.

Instead of the five moral foundations proposed by MFT, MMM suggests there are six moral concerns that can be mapped along two dimensions. The first dimension distinguishes approach from avoidance moral motives: prescriptive calls to provide help versus proscriptive calls forbidding harm. The second dimension distinguishes different contexts to which these motives apply: the self, the other, or the collective. Crossing these dimensions produces three approach-based goals oriented toward helping the self (industriousness), others (helping/fairness), and the collective (social justice), and three avoidance-based goals oriented toward protecting the self (moderation), others (not harming), and the collective (social order).

MMM presents both a new framework for organizing the moral realm and a challenge to MFT’s claim that only conservatives moralize group concerns. MMM argues that conservatives specifically have stronger avoidance group-level concerns: They are proscriptively driven to protect threats to social order within their group by enforcing homogeneity and strict norm adherence. In contrast, liberals have stronger approach group-level concerns: They are prescriptively driven to promote social justice across groups via encouraging interdependence and shared responsibility. Supporting this hypothesis, across two studies, liberal participants more strongly endorsed the importance of providing for communal welfare whereas conservative participants more strongly endorsed social conformity and order. But liberals and conservatives did not differ in their endorsements of interpersonal (as opposed to group-level) concerns, like going out of one’s way to help others or not taking advantage of others (Janoff-Bulman & Carnes, Reference Janoff-Bulman and Carnes2016).

A second alternative account to MFT, the AHA (Gray et al., Reference Gray, MacCormack, Henry, Banks, Schein, Armstrong-Carter, Abrams and Muscatell2022; also see its theoretical ancestor, the theory of dyadic morality; Schein & Gray, Reference Schein and Gray2017), argues that political disagreement arises solely from divergent perceptions of harm. Like MMM, this account challenges MFT’s five distinct moral concerns, suggesting that harm is the fundamental perception involved in all moral judgments: People judge an act as morally wrong when they perceive that an intentional agent has harmed a vulnerable patient (i.e., an entity capable of suffering). To explain MFT’s five-dimensional findings while positing an evolved aversion only to harm – not to injustice, impurity, disloyalty, or disrespect of authority – AHA argues that culture and personal experiences lead groups and individuals to differ in who or what they perceive to be an intentional agent or vulnerable patient. Through this lens, the fundamental moral concerns specified by MFT (or even MMM) are merely descriptive labels that people use to categorize different types of harm in different types of dyads.

Thus, the fundamental conflict between MFT and the AHA hinges on whether people can have intuitive moral judgments in situations where there is no explicit harm nor a harmed victim. While MFT points to moral judgments in scenarios that are ostensibly harm-free (e.g., a person cooking and eating their pet after it dies), AHA argues that such scenarios still involve perceived harm, in this case perhaps to the soul of the pet or its owner. Turning to a political example, when conservatives say it is immoral to do something disloyal like burn one’s country’s flag, AHA assumes they must infer a vulnerable patient who suffers as a result (perhaps their country) and that they endorse the binding moral value of loyalty because of that perceived harm. Supporting this account, both liberals’ and conservatives’ judgments of an action’s (im)morality most closely track their perceptions of how harmful (as opposed to impure, disloyal, unfair, or disobedient) the act is (Schein & Gray, Reference Schein and Gray2015). Likewise, support for anti-GMO policies tracks perceptions of how harmful GMOs are better than how impure they are (Gray & Schein, Reference Gray and Schein2016).

Both MFT and its challengers continue to accrue evidence in support of their respective theories of human moral concerns and how these concerns fuel political disagreements. In responding to each other’s theoretical and empirical challenges, they have evolved over time. For instance, MFT researchers have expanded their moral pantheon to include the liberty foundation, creating new measures to capture concerns over freedom from oppression (Clifford et al., Reference Clifford, Iyengar, Cabeza and Sinnott-Armstrong2015). They find that liberals emphasize this foundation more than conservatives, which helps to account for liberals’ social justice motives highlighted by MMM. They also find that libertarians emphasize this foundation more than liberals or conservatives do (Iyer et al., Reference Iyer, Koleva, Graham, Ditto and Haidt2012), extending moral-political theorizing beyond liberal-conservative dichotomies. While sorting out their disagreements about how morality fuels political beliefs and divides, these accounts have largely overlooked the possibility that political beliefs fuel morality. We consider this possibility next.

22.1.3 Can Political Beliefs Instead Cause Moral Beliefs?

All three accounts we have discussed assume that moral values cause people to adopt particular political beliefs. This position is intuitively plausible: Most people’s introspective experience is that they carefully consult their moral values, and of course the relevant facts, and use those as the basis from which to choose their policy positions. Nevertheless, there is reason to entertain the opposite possibility: That people often know which side of an issue they want to support, and they recruit moral values (and facts) to justify this position. This pattern reflects motivated cognition, which often occurs when people defend their political views (Liu & Ditto, Reference Liu and Ditto2012).

At least some evidence suggests that liberals and conservatives selectively endorse moral values that justify their preferred political positions (Uhlmann et al., Reference Uhlmann, Pizarro, Tannenbaum and Ditto2009). This work examined preferences for moral consequentialism, or the principle of maximizing positive outcomes overall. Conservatives read about a military policy that would ensure the greater good at the cost of some civilian lives, but the researchers manipulated whether those lives would be Iraqi or American. These participants supported the policy more in the former condition and justified their position by more strongly endorsing consequentialist values. Liberal participants faced the prospect of sacrificing one individual’s life to save a hundred, but the researchers manipulated whether that one individual had a stereotypically Black or White American name. These participants preferred not to sacrifice the Black individual and justified their position by denouncing consequentialist values.

That set of studies suggests people’s moral values (in this case, their endorsement of consequentialism) can come from, rather than shape, their political preferences. Providing converging evidence, cross-lagged analyses in three separate panel studies revealed that political ideology better predicts people’s endorsement of the five moral foundations over time than the opposite (Hatemi et al., Reference Hatemi, Crabtree and Smith2019). These data suggest that people’s political allegiances change which moral values they adopt, rather than the other way around. That said, both processes likely coexist and each may dominate at different times.

22.2.1 Moral Concerns Encourage People to Act in Support of Their Favored Policies

In pluralistic democratic societies where people disagree and successful policies require majority support, people can directly help their preferred policies’ chances both through individual efforts (e.g., voting, speaking out, signing petitions) and by successfully inspiring others to join in collective action. When people’s political beliefs are based on moral convictions, they are more likely to undertake these sorts of behaviors, in ways that can be more or less constructive.

22.2.2 Morality Undermines Motivation to Engage Constructively with Political Opponents

As we have seen, moral concerns can drive people to engage in actions aimed at promoting their political goals, though sometimes these actions are less effective and socially sanctioned. But in pluralistic democratic societies, promoting one’s favored policies may not be enough: Achieving majority support for a policy often requires engagement between opposing factions, as opponents can be persuaded to join the cause or negotiated with to at least partially advance it. Though people endorse various moral values that encourage such cross-divide engagement in principle, their other moral concerns (paradoxically) undermine it in practice.

Though there are practical and principled reasons for politically motivated people to engage across political divides, they seldom do so. The same moral values that fuel direct political action also keep people away from their political opponents and out of cross-party conversations that could actually help their political cause. Specifically, moral concerns both pull people toward those with politically similar beliefs, and push them away from those who hold different political opinions.

22.2.3 How People Morally Justify Disengagement from Cross-Ideological Dialogue

If people in principle believe they should be tolerant and wish to make political progress but in practice avoid opponents and even intentionally suppress their views, how do they not see themselves as moral hypocrites? Consider these three explanations.

First, people might not even notice that their segregationist behavior violates their tolerant values and pragmatic political interests (Crawford & Pilanski, Reference Crawford and Pilanski2014). If people unintentionally gravitate toward politically congenial people and information, they may be oblivious to their exclusion of opponents. When they move to a neighborhood that feels right to them, they may not realize this precludes friendly connections with political adversaries. And when they choose the comfort of ideologically aligned news, they may fail to notice how this violates their open-minded values.

Second, people may notice the disconnect but refuse responsibility for it. As noted earlier, people excel at justifying desired conclusions; if partisans are motivated to find fault in their opponents, they may find ways to blame their lack of contact with opponents on the opponents themselves. For example, partisans might claim that they are willing to engage, if only their close-minded opponents were equally willing to socialize (Iyengar et al., Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019; Iyengar & Westwood, Reference Iyengar and Westwood2015). Or they might argue that they have tried to engage with opponents before and therefore know how useless it would be to engage further, perhaps claiming that their opponents are stubbornly immune to persuasion or compromise, or that they already know exactly what they will say (Yeomans, Reference Yeomans2021). These rationalizations blame opponents for the impasse, allowing partisans to acknowledge that they are not engaging without feeling guilty about it.

Finally, people may see their political stonewalling as morally righteous (Hawkins et al., Reference Hawkins, Yudkin, Juan-Torres and Dixon2019). If conservatives view liberals as overly sensitive, unpatriotic flag-burners who welcome criminal immigrants, they likely find it perfectly justified – or even necessary – to sacrifice tolerance at the more sacralized altar of national security. If liberals view conservatives as heartless, gun-brandishing racists, they may similarly forgo tolerance to advance their sacred value of racial justice. Indeed, people perceive prejudice toward moral opponents to be uniquely justified (Cole Wright et al., Reference Cole Wright, Cullum and Schwab2008), feeling no dissonance when they disparage, censor, and disregard groups with dissimilar morals (Crawford & Pilanski, Reference Crawford and Pilanski2014; Iyengar & Westwood, Reference Iyengar and Westwood2015). While people endorse empathy and tolerance in the abstract, they believe these should be withheld from immoral people (Haidt et al., Reference Haidt, Rosenberg and Hom2003; Wang & Todd, Reference Wang and Todd2020).

22.2.4 Improving Motivation to Engage With, and Attitudes Toward, Political Opponents

People actively avoid engaging with their political opponents, so interventions fostering positive cross-party engagement must first overcome this motivational barrier, by making people see positive engagement as a desirable goal. Since most people endorse tolerance, helping them see the hypocrisy of their avoidant behavior might induce them to act differently (Batson et al., Reference Batson, Thompson, Seuferling, Whitney and Strongman1999). Of course, people could instead resolve this hypocrisy by disavowing tolerance; to prevent this, interventions should emphasize tolerance as a primary moral virtue (see Kovacheff et al., Reference Kovacheff, Schwartz, Inbar and Feinberg2018). Alternatively, interventions could leverage social pressures. People are strongly motivated to behave in ways that their group approves of (Hillman et al., Reference Hillman, Fowlie and MacDonald2022); since partisans strongly prefer copartisans who seek to better understand, rather than avoid, opponents (Heltzel & Laurin, Reference Heltzel and Laurin2021), interventions could harness this social approval to motivate political discourse. Such interventions could increase people’s awareness of their avoidance of opposing political views and motivate them to start seeking ways to engage.

Still, even if partisans want to engage across divides, their dislike for opponents may doom their attempts. Partisans dislike their opponents and expect to be disliked in return (Lees & Cikara, Reference Lees and Cikara2020). This may create a self-fulfilling prophecy, whereby partisans enter into a cross-party interaction feeling defensive and unforgiving, causing it to go poorly. That is, because they expect to be disliked, partisans may approach an opposing interlocuter with cold indifference, offending the opponent and thereby creating the chilly atmosphere they initially expected. Such interactions likely only reinforce negative stereotypes of opponents, hampering productive conversation and dissuading well-intentioned partisans from trying again. Interventions seeking to foster productive and sustained cross-divide engagement, then, should not only motivate people to engage but also inoculate against partisan animosity that would otherwise foil pleasant engagement. To this end, we discuss intervention strategies that target individuals’ attitudes and interpersonal relationships (for a broader review, see Hartman et al., Reference Hartman, Blakey, Womick, Bail, Finkel, Han, Sarrouf, Schroeder, Sheeran, Van Bavel, Willer and Gray2022).

22.2.5 Improving Constructiveness of Contact and Dialogue

Once people are motivated to engage in cross-party dialogue and like opponents enough to approach it in good faith, there are still many opportunities for the conversation to derail. We close by identifying conditions under which political conversations, once initiated, remain pleasant and constructive, especially given that conversation may drift to morally relevant (and therefore potentially divisive) topics.

Conversations between partisan opponents may fare better when participants foster empathy by highlighting shared moral ground. Since morality is key to building trust and liking, cross-party conversations will naturally fare better when they invoke shared moral values (e.g., care and fairness; Graham et al., Reference Graham, Haidt and Nosek2009). But even when conversations bring up moral disagreements, partisans can avoid conflict by discussing how their personal experiences have informed their political views: Compared to facts that can be dismissed as fake news, personal experiences are difficult to refute and easier to empathize with (Kubin et al., Reference Kubin, Puryear, Schein and Gray2021), allowing even staunch opponents to see opposing views as reasonable and legitimate (Stanley et al., Reference Stanley, Whitehead, Sinnott-Armstrong and Seli2020). For example, conservatives respect a liberal’s gun control stance more if the stance stems from having suffered from gun violence and less if it stems from statistics supporting gun regulation. These strategies can allow people to talk about the loaded moral beliefs that infuse their political views while still increasing empathy and improving attitudes toward opponents.

Another strategy to facilitate political discussions is to make political opponents feel respected, heard, and included. Respect can make interactions friendlier. Telling political opponents that one respects their status can make them less defensive (Moore-Berg et al., Reference Moore-Berg, Ankori-Karlinsky, Hameiri and Bruneau2020) and more friendly toward you. In one study, when an opponent who disagreed nevertheless acknowledged respect for participants’ views on the Affordable Care Act, participants viewed the opponent more positively and were more willing to give them money in a dictator game (Bendersky, Reference Bendersky2014). Making sure your opponents feel heard can have similar effects (Yeomans et al., Reference Yeomans, Minson, Collins, Chen and Gino2020). This is even true among groups with a history of violence: When members of conflict-affected communities in Colombia were able to tell their personal experiences to ex-combatants on the other side, they liked the ex-combatants more (Ugarriza & Nussio, Reference Ugarriza and Nussio2017). And when people include rather than exclude an opponent, even in low-stakes online conversations, the opponent likes them more and sees them as more moral (Voelkel et al., Reference Voelkel, Ren and Brandt2021).

These strategies may also succeed because they elicit reciprocation, opening the door to good-faith compromise and agreement. When people listen to each other, their opinions shift closer to the center. For example, after door-to-door canvassers nonjudgmentally listened to participants’ personal narratives about immigration policy, those participants shifted their views closer to the canvassers’ (Kalla & Broockman, Reference Kalla and Broockman2020). When a person hears their opponent out, they can better acknowledge their opponent’s arguments, making them seem informed and unbiased (Hussein & Tormala, Reference Hussein and Tormala2021; Xu & Petty, Reference Xu and Petty2021); as a result, opponents may feel more willing to soften their position and compromise. For example, when liberals show they properly understand conservatives’ pragmatic concerns about the economic opportunities of pipeline projects, conservatives may feel more receptive to reasons for canceling pipeline projects. When people have heard each other, they can also speak to each other’s moral concerns, and find common ground in that way; for example, liberals can highlight the purity violations inherent to pollution, and conservatives can highlight the fairness benefits of funding the military as an employer and educator of the disadvantaged (Feinberg & Willer, Reference Feinberg and Willer2013).

23.1.1 Conceptual Space and Terminology

To begin, let us define our central concepts. We treat both “morality” and “religion” systemically, that is, each is comprised of causally interconnected components that produce some output (Meadows, Reference Meadows2008; von Bertalanffy, Reference von Bertalanffy1968). As such, without further specification, we use “morality” and “religion” to mean moral and religious systems, respectively. The general question that drives much of contemporary research is the question of the degree to which these systems overlap and influence each other.

We define religious systems as shared beliefs in spiritual or supernatural agents (e.g., gods and ghosts) or processes (e.g., karma or mana), shared behaviors done with appeals to them (Jensen, Reference Jensen2019; Tylor, Reference Tylor1920; cf. Purzycki & Sosis, Reference Purzycki and Sosis2022), and the dynamic relationship between them. This is a broad and cross-culturally inclusive definition and therefore does not limit the focus to relatively more organized, formalized, dogmatic, or doctrinal systems (cf., Boyer, Reference Boyer2018, p. 21; Sperber, Reference Sperber2018). It also stresses a supernatural element and hence excludes other domains of “ultimate concern” (Tillich, Reference Tillich1957) such as politics, music, and sports from its conceptual space. Moreover, this definition also stresses religious behavior and therefore avoids the ethnocentric assumption of treating religion simply as faith or belief in supernatural agents (see Cohen et al., Reference Cohen, Siegel and Rozin2003; Kavanagh & Jong, Reference Kavanagh and Jong2020).

We define moral systems as ideas and behaviors associated with norms (i.e., content) of interpersonal social behaviors. These behaviors impose benefits (the “good”) or costs (“bad”) on others (Alexander, Reference Alexander1987; Curry et al., Reference Curry, Whitehouse and Mullins2019; Graham et al., Reference Graham, Haidt, Koleva, Motyl, Iyer, Wojcik and Ditto2013; Purzycki, Pisor, et al., Reference Purzycki, Pisor, Apicella, Atkinson, Cohen, Henrich, McNamara, Norenzayan, Willard and Xygalatas2018) and their corollary behaviors (or absence thereof). Moral systems’ constituent parts include: 1) intuitions (i.e., judgments about the “good” and “bad”); 2) reasoning (i.e., the process by which one answers questions like “do I think x is good or bad?” or “which is less immoral, x or y?”); 3) models (e.g., “bad behaviors are x, y, z…”); 4) culture (e.g., “we think x, y, z…” or “population Q thinks x, y, z are bad”); and 5) behavior (i.e., actions judged as morally good or bad). Like our conception of religion, this definition stresses the behavioral aspect of morality and its dynamic relationship with individual and group values (see Box 23.1 and McKay & Whitehouse, Reference McKay and Whitehouse2014, for further discussion).

23.1.2 Some Problems in the Study of Morality and Religion

Contemporary research asks: Where do morality and/or religion come from? Do they relate to each other? If so, how? Answers, of course, come from all quarters. For instance, both believers and nonbelievers embedded in the Abrahamic traditions claim that religion and morality are fundamentally intertwined and that religion is “about” morality (see Abrams, Reference Abrams2022; Dennett Reference Dennett2006, ch. 8), often even claiming that humanity’s collective moral standards were passed down from God or Holy Scripture and that the world would descend into chaos without religion upholding those standards. In active research, the main questions driving the field are: 1) What best accounts for religion/morality: cognitive processes or social learning? 2) Does religion get us to behave in ways that count as “moral”? 3) Are aspects of morality necessary for religious beliefs? 4) What best accounts for religions that are explicitly associated with morality? We address these questions in the present chapter.

23.2.1 Agency-Detection and Religious Beliefs

There is broad scholarly agreement that religious thought and behavior rely on evolved, cognitive foundations that reliably develop in humans. There is less agreement, however, on the specific cognitive mechanisms that constitute these foundations. One mechanism, however, plays a major role in most cognitive and evolutionary accounts of religion, namely, the ability to represent other minds. Humans are equipped with propensities – most likely unrivaled in the animal kingdom – for detecting and inferring the existence and content of other minds (Call & Tomasello, Reference Call and Tomasello2008; Penn & Povinelli, Reference Penn and Povinelli2007; Premack & Woodruff, Reference Premack and Woodruff1978). This so-called theory of mind system, also known as the mentalizing system, allows people to make sense of and predict intentions and behavioral patterns of other individuals (Baron-Cohen, Reference Baron-Cohen1995; Veissière et al., Reference Veissière, Constant, Ramstead, Friston and Kirmayer2020). Presumably, such capabilities have been favored by selective pressures during hominid evolution in a dynamic relationship with increased group size and social complexity, which in turn may have presented both opportunities and challenges for higher-level flexible social cognition, including Machiavellian tactics and sophisticated social learning (Dunbar & Shultz, Reference Dunbar and Shultz2007, Reference Dunbar and Shultz2017; Markov & Markov, Reference Markov and Markov2020; Muthukrishna et al., Reference Muthukrishna, Doebeli, Chudek and Henrich2018; van Schaik & Burkart, Reference van Schaik and Burkart2011; van Schaik et al., Reference van Schaik, Isler and Burkart2012).

Indeed, the human propensity for detecting and inferring other minds appears relatively “oversensitive” as it sometimes extends to domains where minds are not present, such as trees, fluids, and abstractions (Heider & Simmel, Reference Heider and Simmel1944). On this basis, many cognitive and evolutionary approaches to religion assume that beliefs about and acts dedicated to supernatural agents partly originate from these mentalizing capabilities (e.g., Atran & Henrich, Reference Atran and Henrich2010; J. L. Barrett, Reference Barrett2000; J. L. Barrett & Richert, Reference Barrett and Richert2003; Bering, Reference Bering2006; Boyer, Reference Boyer2001; Guthrie, Reference Guthrie1980, Reference Guthrie1995; Norenzayan et al., Reference Norenzayan, Shariff, Gervais, Willard, McNamara, Slingerland and Henrich2016a; Peoples et al., Reference Peoples, Duda and Marlowe2016; for an alternative perspective, see Andersen, Reference Andersen2019). These capabilities allow us to imagine, infer, and articulate the desires and perceptions of spiritual agents. In a similar way, many current researchers approach morality as a set of evolved, foundational cognitive systems. What distinguishes research into these domains is that while virtually no one claims that mentalizing evolved for believing in gods, many suggest that cognitive systems evolved for – or even are – morality (e.g., Baumard, Reference Baumard2016; Greene, Reference Greene2013). We discuss these hypothesized cognitive systems in Section 23.2.2.

23.2.2 Morality as Cognitive Machinery

While the enormity of the evolutionary literature on morality implies its diversity, there are some core themes and elements that cross-cut topics. These stem from the evolution of cooperation literature, which generally asks: Why would an individual engage in costly acts that benefit others? The standard way to begin answering this question is by appealing to kin selection, or the preferential investment in those more closely related to you. In evolutionary terms, individuals are more inclined to engage in costly behavior when it benefits closer relations because it still benefits themselves genetically. This leads to the question: Why would nonkin engage in cooperative behavior toward others if they do not share genes to benefit? The typical response is that some form of “reciprocal altruism” (Axelrod, Reference Axelrod1984; Trivers, Reference Trivers1971) can be adaptive insofar as individuals will punish those who cheat them and reciprocate after others invest in them. Such reciprocity is relatively rare in the natural world, however, and humans are among the most adept at designing ways to ensure various forms of reciprocity.

As many others before us, we illustrate the problem of morality and cooperation using the game-theoretic prisoner’s dilemma, “an abstract formulation of some very common and very interesting situations in which what is best for each person individually leads to mutual defection, whereas everyone would have been better off with mutual cooperation” (Axelrod, Reference Axelrod1984, p. 9). Touted as “the purest expression of the conflict between individual and group interests” (McElreath & Boyd, Reference McElreath and Boyd2008, p. 72), the prisoner’s dilemma clearly points to features of morality and the evolution of cooperation. We therefore use it in this chapter to illustrate how various approaches address the relationship between morality and religion. Table 23.1 represents the basic dilemma in the form of a payoff matrix. Here, b and c refer to benefits and costs, respectively.

In this dilemma, you can either cooperate (C) or defect (D). Assuming we are Player 1, we can only choose between these two options, but what we end up getting depends on what Player 2 does. The best-case scenario for us is defecting when the other player cooperates. If we both defect, there is relatively no gain or loss, and if we cooperate and the other player defects, we are the worst off. In this game, without further specifications like relatedness or reciprocal strategies, it is always better to defect.

Let’s say you have a population of individuals who always cooperate and some who always defect. If p is the proportion of cooperators in the population and 1 − p is the remaining proportion of defectors, then the total payoffs for those who always cooperate is the probability of interacting with a cooperator times the payoff, p(b − c), and the probability of interacting with a defector times that payoff, (1 −p)(−c), for a total payoff of pb − c. The payoffs for individuals who always defect, then, would be pb + (1 − p)0 = pb. As p(b − c) < pb, defectors will always outcompete cooperators. How, then, do we avoid the world engulfed by avarice that this model predicts?

Thankfully, the world in which we live is not this bleak, simplified world (Johnson et al., Reference Johnson, Stopka and Bell2002). We therefore require some account of 1) mechanisms that offset the probability of choosing defection, 2) means to prevent defectors from entering the playing field, 3) strategies that would outcompete defectors, or 4) ways to alter the payoff structure entirely. As already discussed, researchers posit that a variety of mechanisms such as kinship and reciprocity overturn the draw of defection.

Recent movements in the evolutionary psychology of morality explicitly link the problem of cooperation with morality (for critical discussion, see Baumard et al., Reference Baumard, André and Sperber2013; Curry, Reference Curry, Shackelford and Hansen2016). For example, Greene (Reference Greene2013, p. 23) writes: “Morality is a set of psychological adaptations that allow otherwise selfish individuals to reap the benefits of cooperation” and “[t]he essence of morality is altruism, unselfishness, a willingness to pay a personal cost to benefit others.” Another example comes from Baumard (Reference Baumard2016, pp. 72–73) who emphasizes the mutualistic aspects of morality and argues that “the selective pressures that explain the emergence of a moral sense push [individuals] toward considering each person’s interests impartially.” Such approaches focus on mental machinery that increase the likelihood of overcoming selfish behavior. According to these views, then, morality and the “moral sense” are mechanisms that make people cooperate. As we discuss in Section 23.3, some suggest that elements of religions function to galvanize such systems that contribute to increasing the likelihood that individuals will cooperate.

The story becomes a little more complicated when we examine the range of complexity in human societies. While these aforementioned mechanisms generally hold for small-scale social organizations, humans have also created large interconnected social systems of nonkin who cannot possibly directly reciprocate. For some, this fact underscores limitations to classical mechanisms of cooperation; because humans are “hypersocial” and invest in many anonymous individuals who can never reciprocate, some other mechanisms are required to account for this level of sociality. Researchers have proposed a host of mechanisms to account for human levels of costly social behavior, including conflict, punishment, and social norms (e.g., Fehr & Fischbacher, Reference Fehr and Fischbacher2004; Fehr et al., Reference Fehr, Fischbacher and Gächter2002; Richerson & Boyd, Reference Richerson and Boyd1999; Turchin et al., Reference Turchin, Currie, Turner and Gavrilets2013). When it comes to religion, current debates revolve around whether certain traditions can contribute to increased complexity or if certain aspects of religious systems are better thought of as responses to social dilemmas and problems, including those particular to social complexity. As these debates tend to revolve around the persuasion of evolutionary social science of those involved, we briefly situate these views in their varied backgrounds.

23.2.3 Evolutionary Approaches to Social Life

Broadly speaking, there are three general approaches in the evolutionary social sciences (Smith, Reference Smith, Cronk, Chagnon and Irons2000), evolutionary psychology, dual-inheritance, and behavioral ecology. In their pursuit of understanding human behavior and culture, evolutionary psychologists tend to emphasize evolved psychology (see Section 23.2) in its pursuit of understanding human behavior and culture. In the case of religion, the reigning evolutionary psychological view is that while we have cognitive systems selected for moral behavior, religious concepts and behaviors are simply extensions of these and other evolved traits (e.g., Boyer, Reference Boyer2001). Dual-inheritance theory stresses the importance of both genetic and cultural transmission as lifting the bulk of explanatory weight for human social life. In the field of religion and morality, it tends to emphasize norms and institutions and much of the work branded as dual-inheritance in this sense focuses primarily on explaining the aforementioned problem of human hypersociality rather than religion per se (Atran & Henrich, Reference Atran and Henrich2010). In contrast to these approaches, human behavioral ecology focuses on optimal adaptive behaviors that provide benefits for individuals and tends to ignore mental processes and cultural transmission (Shaver & Sosis, Reference Shaver and Sosis2014; Sosis & Bulbulia, Reference Sosis and Bulbulia2011). Arguably, all three of these approaches to understanding cooperation are inherently ecological as they focus on the distribution of energy throughout social systems with an eye toward the cost-benefit trade-offs.

In practice, virtually no one argues that humans have a specifically evolved predisposition for producing religious beliefs and practices. But there are many influential arguments that posit that religious beliefs and practices are easier to learn because they exploit evolved cognitive systems. Content-biased cultural transmission (Richerson & Boyd, Reference Richerson and Boyd2005) is the process by which beliefs or traditions spread and stabilize in a population because they are inherently more “attractive” (Sperber, Reference Sperber1996) than others, as they resonate with deeper evolved psychological systems. For example, moral norms might be easier to learn because violating them is costly or that they resonate with evolved moral cognition. In addition to our social cognitive systems, a range of content biases have been proposed as cognitive foundations for religious thought, including mentalizing. In other words, cultural concepts that activate this mentalizing machinery may enjoy advantages in terms of memorability and transmissibility compared to cultural concepts that do not (Mesoudi et al., Reference Mesoudi, Whiten and Dunbar2006). This enables the genesis and diffusion of mental representations of nonphysical entities with their own wants, needs, knowledge, intentions, etc., such as gods, ghosts, and spirits. In Section 23.3.3, we discuss moral cognition as another potential content bias of spiritual agents.

23.3.1 Supernatural Monitoring, Punishment, and Cooperation

What has become known as the “supernatural punishment hypothesis” (Johnson, Reference Johnson2005, Reference Johnson2016) posits that a suite of mechanisms that contribute to cooperation evolved in competition with manipulative Machiavellian behavior motivated by selfishness. The evolutionary story of supernatural punishment goes as follows. In hominids’ distant past, agency detection (see Section 23.2.1) proliferated as it ostensibly facilitated anticipating the behaviors of other entities. This aptitude might have contributed to our mastery of hunting and propelled our sociality in unprecedented ways. However, once hominids could anticipate others’ mental states, they could also manipulate them for their own benefit at others’ expense. This opens up a new avenue for problems associated with cooperation and coordination. To reduce the appeal of defection, a sensitivity toward commitment to a being that monitors your behavior and can punish you may have evolved. A simple model stipulates the conditions under which such “god-fearing” (i.e., the psychology that supports the deference to dominant agents who are not obviously there) would outcompete Machiavellian strategies. Specifically, it posits that when the probability, p, and costs, c, of getting caught for defecting outweighs the missed benefits, m, that could have been reaped if one defected – that is, when pc > m – god-fearing will evolve. In other words, wherever the costs of punishment and effective monitoring outweigh whatever benefits one would have gotten by being bad, god-fearing will eventually become widespread in a population.

While the supernatural punishment model addresses the contexts in which god-fearing will outcompete self-interested strategies, it takes only a little more effort to appreciate the implications it has on cooperation more generally. Accepting that a god can watch and punish you increases the product of pc. In a social dilemma like the prisoner’s dilemma, this mechanism might propel individuals toward taking the cooperative option (i.e., forgoing m). Appeals to gods’ punishment might also alter the perceived payoffs for defecting. For example, while b < c in Table 23.1, god-fearing might redefine this inequality to b > c, where the benefits of cooperation outweigh the spiritual costs of defecting. We can also achieve this by adding a penalty for defecting (e.g., b − p and −p for the payoffs in the second row of the table). In this case, cooperators would outcompete defectors whenever p > c. Other possibilities abound (see discussion between Johnson, Reference Johnson2011: Lane, Reference Lane2018, Schloss & Murray, Reference Schloss and Murray2011, for a more elaborate simulation), but this example illustrates how supernatural punishment might adaptively avoid such costs of actual punishment.

Assuming there is always some variation between (and within) individuals, we should expect to see that harnessing aspects of god-fearing should reduce self-interested behaviors both situationally and longitudinally. There is considerable evidence showing that the beliefs in and/or primed threat of spiritual observers can alter individual performance in economic game experiments in the predicted manner of reducing selfish behavior, at least among coreligionists (e.g., Lang et al., Reference Lang, Purzycki, Apicella, Atkinson, Bolyanatz, Cohen, Handley, Kundtová Klocová, Lesorogol, Mathew, McNamara, Moya, Placek, Soler, Vardy, Weigel, Willard, Xygalatas, Norenzayan and Henrich2019; McKay et al., Reference McKay, Efferson, Whitehouse and Fehr2011; McNamara & Henrich, Reference McNamara and Henrich2018; McNamara et al., Reference McNamara, Norenzayan and Henrich2016; Piazza et al., Reference Piazza, Bering and Ingram2011; Purzycki et al., Reference Purzycki, McNamara, De Cruz and Nichols2016; Rand et al., Reference Rand, Dreber, Haque, Kane, Nowak and Coakley2014; Shariff & Norenzayan, Reference Shariff and Norenzayan2007, Reference Shariff and Norenzayan2011). This literature tends to focus on whether or not variation in explicit beliefs or priming psychological systems associated with spiritual punishment induces cooperative, “moral” behavior and/or reduces selfish, “immoral” behavior. As we now turn to discuss, considerable research also shows that engaging in ritual behaviors associated with spirits and gods can induce the same effects for both observers and participants.

23.3.2 Religious and Moral Behaviors

Religious ritual is directly implicated in the moral lives of people. For Roy Rappaport (Reference Rappaport1999), ritual “not only brings conventions into being but invests them with morality. Moral dicta are not explicit in all liturgies, but morality, like social contract, is implicit in ritual’s structure” (p. 132). Here, ritual enacts or expresses convention but also endows conventions with the kind of obligatory prescriptions that come with the urgency of ritual participation and the salience of its ordered steps. Engaging in this suite of obligations, then, conveys to participants “that he or she accepts the order encoded in the ritual in which he or she is participating” (Rappaport, Reference Rappaport1994, p. 339, emphasis in original). In this view, as ritual intrinsically contains the mores and rules of social life, engaging in ritual reaffirms this connection and conveys to others that he or she accepts those obligations as well. In other words, when people perform a ritual, they transmit a much wider range of information about conduct in social life than merely their beliefs or that they are performing a ritual. Rather, they are conveying adherence to the greater moral expectations of their community.

Rappaport’s sentiments resonate with an important distinction raised by Teehan (Reference Teehan, Liddle and Shackelford2016), who argues that religions need not be explicitly about morality (e.g., with explicit rules or doctrines about morality and/or with gods believed to care about moral behavior) in order to be morally relevant in a more practical sense. For instance, cross-culturally, ritual participation is something gods are often concerned with (Bendixen & Purzycki, Reference Bendixen and Purzycki2020; Bendixen et al., Reference Bendixen, Apicella, Atkinson, Cohen, Henrich, McNamara, Norenzayan, Willard, Xygalatas and Purzycki2024; Purzycki & McNamara, Reference Purzycki, McNamara, De Cruz and Nichols2016; Swanson, Reference Swanson1960). However, if the association between deities and ritual participation actually increases the frequency and effort with which one participates in rituals, and if – as a growing body of experimental work testifies – communal rituals can play a powerful role in strengthening social bonds, increasing prosociality, and signaling in-group membership (e.g., Fischer et al., Reference Fischer, Callander, Reddish and Bulbulia2013; Purzycki & Sosis, Reference Purzycki and Sosis2022; Reddish et al., Reference Reddish, Bulbulia and Fischer2014; Wiltermuth & Heath, Reference Wiltermuth and Heath2009; Xygalatas et al., Reference Xygalatas, Mitkidis, Fischer, Reddish, Skewes, Geertz, Roepstorff and Bulbulia2013), then beliefs in these deities certainly has some practical moral relevance for human relationships.

Indeed, one such approach builds on Rappaport’s views and reframes them in the context of evolutionary theory. Specifically, some posit that ritual can be a form of costly signaling inasmuch as the costs of rituals reliably convey one’s commitment to the general mores of the community. Others have found that engagement in costly rituals elicits more cooperation in experiments (Soler, Reference Soler2012), cooperative requests in social networks (Power, Reference Power2017a, Reference Power2017b), and predicts longevity of communes (Sosis & Bressler, Reference Sosis and Bressler2003) due to effectively keeping out those who are unwilling to pay ritual costs, a mechanism of which has been experimentally substantiated (Lang et al., Reference Lang, Chvaja, Purzycki, Václavík and Stanĕk2022). Moreover, the presence and intensity of ritual costs have been found to covary with contexts such as territorial disputes and warfare where the temptation to defect or leave the group is relatively high (Sosis et al., Reference Sosis, Kress and Boster2007). One stream of research finds that those who engage in rituals are perceived as more trustworthy generally (Purzycki & Arakchaa, Reference Purzycki and Arakchaa2013; Sosis, Reference Sosis2005; Tan & Vogel, Reference Tan and Vogel2008).

While signaling trustworthiness is an effect of participation in rituals, gods are also interested in a host of behaviors other than ritual. Yet, the behaviors they care about do appear to refer to behaviors associated with social dilemmas (Bendixen & Purzycki, Reference Bendixen and Purzycki2020; Purzycki & Sosis, Reference Purzycki and Sosis2022; Purzycki, Bendixen, et al., Reference Purzycki, Bendixen, Lightner and Sosis2022). Take, for example, socially consuming alcohol. Refusing to drink with a friend who drinks might entail some social costs, while drinking together is often perceived as entailing a lot more benefits. In the Quran (Surah Al-Baqarah 2:219), drinking is explicitly framed as entailing more harm than benefits. Institutionalizing this Quranic passage and outright banning alcohol changes the payoff structure to make social drinking much riskier and more costly (Purzycki, Bendixen, et al., Reference Purzycki, Bendixen, Lightner and Sosis2022). In other contexts, spirits are angered by over-exploiting resources and, in some cases, entire regions are off-limits to human use by virtue of their association with the gods (see Purzycki, Reference Purzycki2011; S. Singh et al., Reference Singh, Youssouf, Malik and Bussmann2017). In such cases, it might be tempting to hunt or take plant materials from such areas, even though persistent exploitation might lead to devastation. Populating a region with gods might suffice to maintain local biodiversity and thus sustain a mobile resource. To the extent that living up to these regulations conveys strong moral character, such cases further illustrate the close relationship between religion and morality.

23.3.3 Explaining the Ubiquity of “Moralistic Religious” Beliefs

While Section 23.3.2 attended to the view that it is partly by virtue of religious beliefs and/or behaviors that humans are as cooperative as they are, other research focuses on a particular form of religion. These are the so-called moralistic religions, often treated in apposition to otherwise non- or sub-moralistic religions. Indeed, recent decades have seen a flurry of scientific attention to this particular construct, yet few researchers are clear on the critical questions that framed the current chapter.

One view offers that such traditions are extensions of certain reproductive strategies exerted on class-structured societies. Appealing to a popular – but not unproblematic (see Baldini, Reference Baldini2015; Nettle & Frankenhuis, Reference Nettle and Frankenhuis2020; Sear, Reference Sear2020) – view of “life history theory,” some (Baumard & Boyer, Reference Baumard, André and Sperber2013; Baumard et al., Reference Baumard, Hyafil, Morris and Boyer2015) have suggested that religions become explicitly associated with morality primarily because wealthy elites who opt for having fewer but higher-quality children “moralize” the behaviors of the poorer sectors of society who have more and lower-quality children. Others (Purzycki, Ross, et al., Reference Purzycki, Ross, Apicella, Atkinson, Cohen, McNamara, Willard, Xygalatas, Norenzayan and Henrich2018) show that at the individual level (i.e., not the tradition or society level), despite there being an association between food security and number of children, there is no obvious relationship between commitment to moralistic traditions – that is, the degree to which individuals claim their gods care about morality – and food security. In this case, one group of researchers treats the concept of “moralistic traditions” as a traditional-level property, whereas others treat the concept as an individual-level variable.

Another debate revolves around the relationship between social complexity and so-called moralistic religions (Purzycki & McKay, Reference Purzycki, McKay, Purzycki and Bendixen2023). For decades, anthropologists investigated whether or not small-scale societies had gods that cared about how people treated each other or functioned as moral models to which individuals should aspire (Rappaport, Reference Rappaport1979; Tylor, Reference Tylor1920). Largely due to the findings of anthropological fieldwork, Evans-Pritchard (Reference Evans-Pritchard1965) rendered the debate moribund. Some cross-cultural data sets (Boehm, Reference Boehm, Bulbulia, Sosis, Harris, Genet and Wyman2008; Swanson, Reference Swanson1960) show that evidence of supernatural sanctions for immoral behavior is abundant in the ethnographic literature (see Figure 23.1; for more detailed analyses, see Lightner et al., Reference Lightner, Bendixen and Purzyckiin press). A crude analysis of Swanson’s data shows that ethnographies of societies with less than 50 people have a 51 percent chance of mentioning moralistic supernatural punishment; those between 50 and 399 have a 68 percent chance of mentioning moralistic spiritual punishment. Yet, the current mainstream view maintains that small-scale societies lacked gods that cared about morality or had traditions that were about morality (e.g., Baumard & Boyer, Reference Baumard, André and Sperber2013; Norenzayan, Reference Norenzayan2013; cf. Beheim et al., Reference Beheim, Atkinson, Bulbulia, Gervais, Gray, Henrich, Lang, Monroe, Muthukrishna, Norenzayan, Purzycki, Shariff, Slingerland, Spicer and Willard2019). Considering the first attempt (Purzycki, Reference Purzycki2011, Reference Purzycki2013) to systematically and directly ask people how much they thought their traditional gods cared about and punished people for immoral conduct appeared only a decade ago (for more recent reports, see Bendixen et al., Reference Bendixen, Apicella, Atkinson, Cohen, Henrich, McNamara, Norenzayan, Willard, Xygalatas and Purzycki2024; Purzycki, Willard, et al., Reference Purzycki, Bendixen, Lightner and Sosis2022; M. Singh et al., Reference Singh, Kaptchuck and Henrich2021; Townsend et al., Reference Townsend, Aktipis, Balliet and Cronk2020), we wager that any strong conclusions are at best premature (Bendixen et al., Reference Bendixen, Lightner, Purzycki, Tehrani, Kendal and Kendal2023).

Figure 23.1 Reported presence of moralistic supernatural punishment across society size (left) and probability (logistic transformations of posterior estimates) of selected ethnographies reporting moralistic supernatural punishment with 95 percent credible intervals (right). Data are from Swanson (Reference Swanson1960). Proportions are of each category on x-axis of left panel. Types of moralistic supernatural punishment include 1) health-related punishments, 2) punishments in the afterlife, and 3) unspecified “other.” If present across any of these variables, a society got a score of MSP = 1. Society size categories are as follows: 0 = 1–49 people (n = 17); 1 = 50–399 (n = 13); 2 = 400–9,999 (n = 9); and 3 = ≥10,000 (n = 10). Data and code can be accessed here: https://gist.github.com/bgpurzycki/4edc36a10a3d1ff4e6035a6ab463cee2.

23.3.4 Skepticism about the Relationship between Morality and Religion

Virtually all of these approaches posit that in some way or another, the relationship between morality and religion is both causal and measurable. However, some strains of thought deny that the relationship is as significant or informative as such studies suggest. For example, one idea in the literature is that spirit concepts are generally easier to retain and transmit partly because they are intuitively endowed with “socially strategic information” (Boyer, Reference Boyer2001; Purzycki et al., Reference Purzycki, Finkel, Shaver, Wales, Cohen and Sosis2012). In other words, because people portray their gods and other spiritual agents as interested in some aspect of our behavior, they are intuitively associated with moral concerns and this makes such concepts salient and therefore increases their perceived importance and transmissibility (i.e., content-biased transmission). There is some evidence that spiritual agents are intuitively associated with moral information (Purzycki et al., Reference Purzycki, Finkel, Shaver, Wales, Cohen and Sosis2012; Purzycki, Willard, et al., Reference Purzycki, Willard, Klocová, Apicella, Atkinson, Bolyanatz, Cohen, Handley, Henrich, Lang, Lesorogol, Mathew, McNamara, Moya, Norenzayan, Placek, Soler, Weigel, Xygalatas and Ross2022) and some evidence that socially relevant information is easier to retain, even in the context of experiments requiring participants to remember religious-like stimuli (e.g., Beebe & Duffy, Reference Beebe and Duffy2020; Swan & Halberstadt, Reference Swan and Halberstadt2019, Reference Swan and Halberstadt2020). However, Boyer contends that this association between moral domains and religion is limited to this intuitiveness and that religion itself does not contribute to behavior that we might construe as “moral.” A challenge to this view, though, is evidence indicating that as triggered by spiritual agentic entities, the intuitiveness of moral cognition can alter behavior. For example, Piazza et al. (Reference Piazza, Bering and Ingram2011) told a treatment group of children that a spirit by the name of “Princess Alice” frequented the lab. Despite not having explicitly told the children that Princess Alice cares about or will punish them for misbehavior, these children were less likely to cheat in a virtually impossible game than the control group. In other words, the perception of god-like agents with underspecified moral interests can alter the kinds of behavior that count as “moral.”

Others call for greater conceptual and methodological precision in disentangling the possible relationships between religion and moral behavior (e.g., Bloom, Reference Bloom2012; McKay & Whitehouse, Reference McKay and Whitehouse2014; Teehan, Reference Teehan, Liddle and Shackelford2016). For instance, Galen (Reference Galen2012) systematically reviewed the literature on “religious prosociality” across both naturalistic settings and various laboratory experimental paradigms, such as behavioral economic studies and priming. Among a long list of critical remarks, a key conclusion is that, while a body of work does indeed find evidence that religious people are more prosocial (e.g., in terms of cooperation, generosity, and sharing), this effect diminishes drastically – or is reversed – when the focal recipient does not share religious identity with the participants. Therefore, a distinction must be drawn between universal and parochial religious prosociality, at minimum (see also, e.g., Graham & Haidt, Reference Graham and Haidt2010; Lang et al., Reference Lang, Purzycki, Apicella, Atkinson, Bolyanatz, Cohen, Handley, Kundtová Klocová, Lesorogol, Mathew, McNamara, Moya, Placek, Soler, Vardy, Weigel, Willard, Xygalatas, Norenzayan and Henrich2019; Norenzayan et al., Reference Norenzayan, Shariff, Gervais, Willard, McNamara, Slingerland and Henrich2016b), which in turn calls for serious consideration of contextual factors in doing such studies. Galen (Reference Galen2012) also argues that since much work depends on self-report the literature is confounded by well-known problems with self-report measures such as demand characteristics, social desirability, and stereotype effects. In sum, some researchers maintain that the connection between religion and morality is less straightforward than often thought – or at least, in Galen’s case, that much research on this topic is so riddled with contradictions that improving conceptual clarity and consistency is of utmost importance moving forward (see Box 23.1).

24.3.1 Process Debunking

Process debunking arguments attempt to show that given what we know about the processes underlying moral cognition, we have good reason to think that it will not be trustworthy in many circumstances. The two most prominent examples of this type are evolutionary debunking and psychological process arguments against deontological moral judgments.

Evolutionary debunking arguments (for a review, see Vavova, Reference Vavova2015) have two major premises. First, evolution shaped the moral judgments we make today. Most people, for example, believe that harming other people and breaking promises are wrong, and that parents have a greater obligation to their children than to other people. Evolutionary biology provides a powerful explanation of this fact: Individuals with such moral beliefs were more likely to survive and reproduce; for example, because it enabled them to cooperate more efficiently (Curry, Reference Curry, Shackelford and Hansen2016; Machery & Mallon, Reference Machery, Mallon and Doris2010).

Second, there is reason to doubt that evolution would have resulted in our being able to accurately track mind-independent moral facts – that is, moral facts that are true or false independently of what anyone thinks, believes, or feels about them. Standard evolutionary origin stories of human moral judgment appeal to the pressures of natural selection, which favor mental faculties that enhance biological fitness (FitzPatrick, Reference FitzPatrick and Zalta2016). Yet it is hard to see how the ability to track mind-independent moral facts accurately would have benefited biological fitness.

Evolutionary debunkers conclude that our moral judgments will frequently be off-track (i.e, not accurately track moral facts). Some have argued that this undermines versions of moral realism that claim that there are mind-independent moral facts to begin with (Street, Reference Street2006). Other evolutionary debunkers think that we should stop trusting our moral judgments all together (Joyce, Reference Joyce2006).

The other kind of debunking argument – which cites current psychological processes to cast doubt on deontological moral judgments – has been most forcefully made by Greene (Reference Greene and Sinnott-Armstrong2008; also Singer, Reference Singer2005). According to Greene, moral judgments are the output of two distinct types of processing, System 1 and System 2. System 1 processes are unconscious, automatic, fast, and intuitive, while System 2 processes are conscious, controlled, slow, and effortful. The two types of processing often produce conflicting moral judgments. In particular, System 1 tends to output deontological judgments (supported by appeals to rights and duties). In contrast, System 2 tends to output consequentialist judgments (supported by cost-benefit considerations).

Greene thinks that this model has important implications for the trustworthiness of deontological moral judgments. He argues that because System 1 processes are unconscious, fast, and automatic, they are often blunt, inflexible, and unresponsive to relevant evidence. Therefore, we should rely on System 1 outputs only if they have been sufficiently shaped through evolutionary, cultural, or personal experience. However, according to Greene, we have had inadequate experience with many moral problems, especially moral problems that arise in the modern world. Hence, we should not rely on System 1 to make judgments about such problems. And because System 1 tends to output deontological moral judgments, this means that we should not trust our deontological moral judgments in many circumstances.

Both of these process debunking arguments have received a lot of attention in the literature (e.g., Nichols, Reference Nichols2014; Sauer, Reference Sauer2018), including considerable push-back (e.g., Berker, Reference Berker2009; Kahane, Reference Kahane2011; Chapter 5, this volume). Whether or not such arguments succeed remains controversial.

24.3.2 Argument from Disagreement

The second type of argument from empirical premises against the trustworthiness of moral judgment is based on cases of moral disagreement (Tersman, Reference Tersman, Zalta and Nodelman2022). Moral disagreements occur when different individuals or groups of individuals make conflicting moral judgments about an issue, problem, or scenario.

Some instances of moral disagreement are easily resolved. For example, when young children disagree with their parents about moral issues, most of us will side with the adult’s judgment over that of the child. Many philosophers think that we are justified in doing so, because there are relevant differences between the disagreeing sides. For example, the cognitive abilities of young children are usually less developed than those of their parents. Other differences that can help us determine which side of a moral disagreement is more likely to be correct include differences in background knowledge, the amount and quality of evidence brought to bear on the issue, and psychological biases (for a more complete list, see Frances & Matheson, Reference Frances, Matheson and Zalta2019).

Moral disagreements between young children and their parents are easily resolved because it is usually not hard to find epistemically relevant differences between the disagreeing sides. However, there are moral disagreements where this is more difficult. One set of examples like this are moral disagreements that arise between cultural groups (Graham et al., Reference Graham, Meindl, Beall, Johnson and Zhang2016; Chapter 20, this volume). Other examples are due to demographic differences in moral judgment, including gender, age, socioeconomic status (all Table 24.1), and religion (Norenzayan, Reference Norenzayan2013). Doris and Plakias (Reference Doris, Plakias and Sinnott-Armstrong2008), for instance, have suggested that differences between Southerners and Northerners in the United States in attitudes toward violence in response to violations of honor (Cohen & Nisbett, Reference Cohen and Nisbett1994) cannot easily be explained by one side being less rational or more biased than the other (for other examples, see Fraser & Hauser, Reference Fraser and Hauser2010; Machery et al., Reference Machery, Kelly and Stich2005).

If this is right, then for many instances of moral disagreements between demographic and cultural groups, the disagreeing sides are in an equally good epistemic position to judge the issue, problem, or scenario. But many have argued that we should not trust judgments about which such epistemic peers disagree (Frances & Matheson, Reference Frances, Matheson and Zalta2019). And since every one of us belongs to some demographic and cultural groups, this suggests that we should not trust our own moral judgments on these disputed issues, problems, or scenarios, either.

Not everyone has been convinced by this argument. Some dispute that to mistrust all judgments involved is the right way to deal with cases of peer disagreement (for an overview, see Frances & Matheson, Reference Frances, Matheson and Zalta2019). For example, some authors have argued that people should in general assign more evidential weight to their own experiences than to the experiences of others (e.g., Huemer, Reference Huemer and Dougherty2011). Others appeal to a notion of self-trust and argue that because people are justified in trusting in their own judgments and the mental faculties that produce them, the mere fact of peer disagreement may speak against the epistemic reliability of the person they disagree with (e.g., Enoch, Reference Enoch2010). Both lines of argument have the same upshot: It can sometimes be reasonable for someone to stick to their own moral judgment over that of their epistemic peer.

Others question whether the moral disagreements we have been citing are really about anything moral at all. They suggest that instead, many (perhaps all) such disagreements are really disagreements about (nonmoral) facts: “[C]areful philosophical examination will reveal … that agreement on nonmoral issues would eliminate almost all disagreement about the sorts of moral issues which arise in ordinary moral practice” (Boyd, Reference Boyd and Sayre-McCord1988, p. 213). As with process debunking arguments, then, it remains an open question whether and to what extent arguments from moral disagreement succeed.

24.3.3 Arguments from Irrelevant Influences

The third type of argument from empirical premises against the trustworthiness of moral judgment cites irrelevant influences on moral judgments. Many things can influence moral judgments, some of which are not at all problematic. For example, people sometimes change their mind about a moral judgment in light of compelling counterarguments (Bloom, Reference Bloom2010; Paxton et al., Reference Paxton, Ungar and Greene2012). Most would agree that hearing compelling counterarguments is a good reason to change one’s mind, that they are relevant to the moral judgments a person should make. Hence, their influence is perfectly legitimate.

However, other influences are less welcome. In particular, for some of the things that influence moral judgments, it is difficult to see how they could be relevant to whether the issue, problem, or scenario that is being judged is morally right or wrong. Some proposed examples of this are framing effects (Nadelhoffer & Feltz, Reference Nadelhoffer and Feltz2008; Sinnott-Armstrong, Reference Sinnott-Armstrong and Sinnott-Armstrong2008a). A moral judgment about an action (or person or problem) is subject to a framing effect if it changes because of morally irrelevant differences in the way that action (or person or problem) is presented (Demaree-Cotton, Reference Demaree-Cotton2016). Prominent examples of framing effects are order effects and word effects. Order framing effects result from multiple scenarios being presented in different orders. Word framing effects result from different but morally equivalent language being used to describe the same moral scenario. Table 24.2 samples the literature on these two types of framing effect.

To see how framing effects could pose a problem for the trustworthiness of moral judgments, it helps to think of them in terms of disagreement within single individuals. When someone’s moral judgment is influenced by framing effects, they are (or would be) making different (incompatible) moral judgments about an issue, problem, or scenario at different times. They are, in other words, disagreeing (or would disagree) with themselves at another time. As we have seen in Section 24.3.2, moral disagreement can pose a challenge to the trustworthiness of the disputed judgments if there are no epistemically relevant differences between the disagreeing sides. Yet plausibly, that a scenario was presented using different wordings or that the same scenarios were presented in different orders (cf., Horne & Livengood, Reference Horne and Livengood2017) is not an epistemically relevant difference. If this is right, then the influence of framing effects on moral judgment reveals cases of peer moral disagreement within the same individual. Thus, if we follow the advice of many epistemologists, we should not trust moral judgments to the extent that they are influenced by framing effects.

Another candidate example of irrelevant influences on moral judgment is social conformity (for a review, see Chituc & Sinnott-Armstrong, Reference Chituc and Sinnott-Armstrong2020). A social conformity effect occurs when a moral judgment is influenced by judgments that others expressed. It would not be a problem if people did not believe the judgments that they stated but uttered them only in order to avoid social conflict. However, that seems unlikely in several experiments because conformity effects also occur online without any interpersonal interaction (Kelly et al., Reference Kelly, Ngo, Chituc, Huettel and Sinnott-Armstrong2017) and because experimental participants continue to express the same judgments when they are alone after leaving the presence of the others whose judgments influenced them (Aramovich et al., Reference Aramovich, Lytle and Skitka2012). Social conformity effects thus suggest that people sometimes change from one moral judgment to an incompatible judgment that they hear others endorse.

Moral philosophers have traditionally emphasized the autonomy of moral judgment (e.g., Aristotle, 2019; Kant, Reference Kant and Hansen1785/1998). If they were right to do so, then social conformity with others’ moral judgments falls short of this philosophical ideal of autonomous moral judgment. Moreover, whether the person making a moral judgment happens to hear from one group instead of another group is not relevant to whether the issue, problem, or scenario that is being judged is morally right or wrong (or good or bad, etc.). If Kelly believes an act is wrong while she is surrounded by people who say it is wrong, but she believes the same act is not wrong while she is surrounded by people who say that it is not wrong, then her moral judgments at these different times cannot both be correct. She still might be justified in trusting one group over the other if she has some independent reason to believe that one group is more likely (or that the other group is less likely) to be correct on this matter. In the absence of any such reason to favor one group, however, she has no reason to favor either moral judgment, so it is hard to see how social conformity by itself could track the truth or produce trustworthy moral judgments in such cases.

For a final example, consider the influence on moral judgment of incidental affect. An affective state (mood or emotion) is incidental when it is unrelated to an issue, scenario, or problem about which a (moral) judgment is made. Anger at injustice is not incidental when it is based on the moral wrongness of the injustice. In contrast, suppose Jordan is in a bad mood because she had a frustrating day at work. On her way home, someone stops her and asks her to donate to Save the Children. But for her bad mood, Jordan would have happily donated; however, today, she does not. In this example, Jordan’s bad mood (due to her frustrating day at work) is completely unrelated to the moral issue she is considering (whether she should donate to Save the Children). Nevertheless, it influences her moral judgment: Had it not been for her bad mood, Jordan would have believed that she should donate.

Many see this kind of influence as problematic (e.g., Cameron et al., Reference Cameron, Payne and Doris2013; Kumar & May, Reference Kumar, May, Suikkanen and Kauppinen2019). Because the source of incidental affective states is completely unrelated to the moral issue, problem, or scenario about which a moral judgment is made, such affective states do not provide any good reason to change one’s moral judgment. Instead, their effect on moral judgment is distorting. Table 24.3 samples the literature.

How does incidental emotion threaten the trustworthiness of moral judgment? Again, it helps to think about such cases as disagreements within single individuals over time. When someone’s moral judgment is influenced by incidental emotion they make (or would make) different (even incompatible) moral judgments about an issue, problem, or scenario at different times. Unlike for framing effects (and more like social conformity), we can easily identify one epistemically relevant difference between the two sides. If incidental emotion distorts moral judgment, then we have reason to discount that side’s judgment. Such disagreements are not like peer disagreements, and consequently do not threaten the trustworthiness of moral judgment in this way.

Even if most moral disagreements due to the influence of incidental emotion can be resolved, however, this does not mean that moral judgments so affected are always home free. It is not enough that in fact a moral judgment was not distorted by incidental emotion. In addition, individuals need to know or at least have good enough reason to believe that a certain moral judgment is not distorted by incidental emotion in order for that individual to be justified in trusting that moral judgment. This might not seem too difficult. Yet adults often have limited insight into the psychological processes underlying their judgment and decision making (Nisbett & Wilson, Reference Nisbett and Wilson1977; Wilson & Dunn, Reference Wilson and Dunn2004), including whether they were influenced by incidental emotion (Schwarz, Reference Schwarz, Van Lange, Kruglanski and Higgins2012). There is considerable reason to think that this is true for the domain of moral judgment specifically (Carlsmith, Reference Carlsmith2008; Greene, Reference Greene and Sinnott-Armstrong2008; Haidt, Reference Haidt2001; Hall et al., Reference Hall, Johansson and Strandberg2012; Hauser et al., Reference Hauser, Cushman, Young, Kang-Xing Jin and Mikhail2007). If this is right, then people will often not be able to tell whether their moral judgments were distorted by incidental emotions. In such cases, the influence of incidental emotion will still make those moral judgments untrustworthy.

24.4.1 Moral Intuitionism

A central epistemological tenet of modern moral intuitionism is that at least some moral judgments are justified noninferentially – that is, justified even in the absence of independent confirmation (Audi, Reference Audi2008; Huemer, Reference Huemer2005; Shafer-Landau, Reference Shafer-Landau2003; cf. Tropman, Reference Tropman2011). According to critics, however, if our moral judgments are often enough not trustworthy, then a given moral judgment would seem not to be justified without at least some independent confirmation (Nadelhoffer & Feltz, Reference Nadelhoffer and Feltz2008; Sinnott-Armstrong, Reference Sinnott-Armstrong and Sinnott-Armstrong2008a).

For an analogy, consider a box of 100 thermometers. You know that many (say, 50) of the thermometers do not show the temperature accurately. You take one of the thermometers out of the box at random. Are you justified in trusting what this particular thermometer says? The answer seems to be “no.” Unless you have some independent reason to think that this particular thermometer is accurate, the fact that many of the thermometers in the box are not accurate should make you skeptical of this one, too. This point applies no matter which thermometer you pick out of the box, so all of them are subject to the same doubts.

If this is correct, then, by analogy, if enough moral judgments are untrustworthy, then no moral judgment should be trusted without independent confirmation. However, that conclusion conflicts directly with the central claim made by moral intuitionists that moral judgments are justified noninferentially, since independent confirmation requires inference. Thus, process debunking, the argument from disagreement, and the argument from irrelevant influences pose serious challenges to moral intuitionism, provided they can show that a wide enough variety of moral judgments are untrustworthy.

24.4.2 How Should We Do Moral Philosophy?

It is widely accepted that one of the main ways (perhaps the main way) to do moral philosophy is to use responses to particular moral problems, issues, or scenarios (Deutsch, Reference Deutsch2010; Williamson, Reference Williamson2007). Such responses have been called the “data of ethics” (Ross, 1930/Reference Ross and Stratton-Lake2002, p. 41). The most widely used method that relies on judgments of particular cases is reflective equilibrium (Kamm, Reference Kamm1993; Rawls, Reference Rawls1971). It consists in “working back and forth among our considered judgments … about particular instances or cases, the principles or rules that we believe govern them, and the theoretical considerations that we believe bear on accepting these considered judgments, principles, or rules, revising any of these elements wherever necessary in order to achieve an acceptable coherence among them” (Daniels, Reference Daniels and Zalta2020).

The problem for this method that is posed by process debunking, the argument from disagreement, and the argument from irrelevant influences is then straightforward: If our moral judgments are too often not trustworthy, then they do not make for a solid foundation on top of which to build moral theory (e.g., Alexander et al., Reference Alexander, Mallon, Weinberg, Knobe and Nichols2014; Horowitz, Reference Horowitz1998; Paulo, Reference Paulo2020). As Jesus is reported to have said (Matthew, 7:26), only a foolish man would build his house on sand. In order to have a secure foundation, moral philosophers who rely on their moral judgments about particular cases would thus need to reconsider the way they build their houses. Of course, defenders of the reflective equilibrium method can respond in several ways, some of which will be discussed in Section 24.5.

24.5.1 Can We Trust the Research?

The argument from disagreement, arguments from irrelevant influences, and some process debunking arguments all rely heavily on findings from moral psychology. Many of these findings have been criticized. Take, for example, the influence on moral judgment of incidental disgust (Table 24.3). Several high-profile studies in this literature have failed to replicate (Ghelfi et al., Reference Ghelfi, Christopherson, Urry, Lenne, Legate, Fischer, Wagemans, Wiggins, Barrett, Bornstein, de Haan, Guberman, Issa, Kim, Na, O’Brien, Paulk, Peck, Sashihara and Sullivan2020; Johnson et al., Reference Johnson, Wortman, Cheung, Hein, Lucas, Donnellan, Ebersole and Narr2016). Moreover, a meta-analysis (Landy & Goodwin, Reference Landy and Goodwin2015) found only a small overall effect, which disappeared entirely once publication bias was taken into account. At this point, it is therefore unclear whether incidental disgust has much, if any, influence on moral judgment.

Null results incompatible with other findings have been published (because of publication bias, there may be more out there that have not been published; Song et al., Reference Song, Parekh, Hooper, Loke, Ford, Sutton, Hing, Kwok, Pang and Harvey2010). For example, Gawronski et al. (Reference Gawronski, Conway, Armstrong, Friesdorf and Hütter2018) failed to find any effect of incidental anger on moral judgments of sacrificial dilemmas. Likewise, several studies have failed to find (or only found negligible) effects of gender, age, and religion on moral judgment (Banerjee et al., Reference Banerjee, Huebner and Hauser2010; Gleichgerrcht & Young, Reference Gleichgerrcht and Young2013; Hauser et al., Reference Hauser, Cushman, Young, Kang-Xing Jin and Mikhail2007).

In addition, some of the studies reviewed exhibit methodological weaknesses, such as low power (Stanley et al., Reference Stanley, Carter and Doucouliagos2018), unrepresentative samples (Henrich et al., Reference Henrich, Heine and Norenzayan2010), and a failure to sufficiently motivate their research question(s) by theory (Muthukrishna & Henrich, Reference Muthukrishna and Henrich2019). Moreover, there is some evidence of questionable research practices in experimental moral psychology more broadly (Stuart et al., Reference Stuart, Colaço and Machery2019), and of publication bias for some of the research programs reviewed in Section 24.3 in particular (Landy & Goodwin, Reference Landy and Goodwin2015; McDonald et al., Reference McDonald, Graves, Yin, Weese and Sinnott-Armstrong2021). Thus, even among results that have not yet been challenged, there is a good chance that some of them will not hold up in the future (Ioannidis, Reference Ioannidis2005, Reference Ioannidis2014).

Proponents of arguments that rely on this research and research like it need to be wary of these scientific issues and should adjust their confidence in their conclusions accordingly. In particular, this means that before we can robustly evaluate the empirical premises of the arguments we surveyed in Section 24.3, we need a lot more – and more rigorous – work in moral psychology (for some recommendations, see Asendorpf et al., Reference Asendorpf, Conner, Fruyt, Houwer, Denissen, Fiedler, Fiedler, Funder, Klieg, Nosek, Perugini, Roberts, Schmitt, van Aken, Weber and Wicherts2013; Shrout & Rodgers, Reference Shrout and Rodgers2018).

24.5.2 Does the Research Show Severe Enough Disagreements?

Arguments from disagreement and arguments from irrelevant influences both cite evidence of moral disagreement (interpersonal in the first case, intrapersonal in the second case). Yet disagreements can be more or less severe. Some disagreements involve moral judgments that are polar opposites – one side judges p, the other side judges not-p. Other disagreements involve moral judgments that do not differ in their polarity but only in their strength or confidence. For example, we might disagree because I think that destroying the environment is extremely wrong, while you think that it is only somewhat wrong.

We can ask whether peer disagreement to any degree counts against the trustworthiness of our moral judgments (Rehren & Sinnott-Armstrong, Reference Rehren and Sinnott-Armstrong2021). Some have suggested that only disagreements involving moral judgments that are polar opposites pose a problem for the trustworthiness of moral judgments (cf. Demaree-Cotton, Reference Demaree-Cotton2016; Kumar & May, Reference Kumar, May, Suikkanen and Kauppinen2019). The thought seems to be that when people (or the same person at different times) vary in the strength or confidence of a moral judgment, then they are still making the same moral judgment (in some sense), and so they are not really disagreeing. If so, disagreements like this involve no real threat to the trustworthiness of the differing moral judgments. On the other side, Andow (Reference Andow2016) has argued that evidence of moral judgments that differ only in strength or confidence can still cast serious doubt on their trustworthiness.

Which side of this dispute one takes has important implications for which kinds of evidence have force. Most of the empirical studies that we have cited throughout Sections 24.3.2 and 24.3.3 provide evidence of one of two forms. Some studies focus on judgment reversals from a moral judgment to its polar opposite. For example, a study that manipulates the order of two scenarios would find evidence of judgment reversals if its participants judged an act to be morally acceptable in one order but unacceptable in the other (e.g., Lanteri et al., Reference Lanteri, Chelini and Rizzello2008). Other studies instead focus on judgment shifts, which occur whenever participants change (or would change) their moral judgments in any significant way, even if only in strength or confidence. For example, the mean response of participants who read two scenarios in one order might be that an act is definitely wrong, while the other order might lead participants to respond only that the act is somewhat wrong (e.g., Wiegmann & Waldmann, Reference Wiegmann and Waldmann2014).

For those who think that peer disagreement provides reason to mistrust the different moral judgments even if this disagreement is only about confidence or strength, but not polarity, evidence of both judgment shifts and judgment reversals is relevant. In contrast, someone who thinks that only peer disagreements that involve polar opposite moral judgments speak against the trustworthiness of our moral judgments should only appeal to studies that provide evidence of judgment reversals. Their evidential basis will therefore be quite a bit thinner.

24.5.3 How Much Untrustworthiness Is Too Much?

Some process debunking arguments (most prominently evolutionary debunking arguments) are global. That is, if they go through, then they implicate all moral judgments. In contrast, the arguments from disagreement and irrelevant influences are local. All they claim to show is that moral judgments should not be trusted to the extent that they feature in intractable moral disagreements or are susceptible to irrelevant influences.

Suppose they succeed: If there is robust evidence that a moral judgment features in intractable moral disagreements or is susceptible to irrelevant influences, then it (and moral judgments like it) is untrustworthy. On its own, this is not enough to discredit moral intuitionism and methods of moral philosophy. It will not be enough to point to just any amount of untrustworthiness, no matter how small. It is clearly unreasonable to demand that moral judgments never go wrong (Shafer-Landau, Reference Shafer-Landau and Sinnott-Armstrong2008).

How much untrustworthiness would be too much? One way to approach this question is in terms of the variety of moral judgments implicated by moral disagreement and irrelevant influences. If this variety is too limited, then the arguments against moral intuitionism and the methods of moral philosophy are weak. Both arguments require that our moral judgments are often untrustworthy. Because people make moral judgments about a wide variety of problems, issues, and scenarios, this requirement can be met only if a large enough fraction of this variety of moral judgments is implicated by moral disagreement or irrelevant influences.

Unfortunately, the existing research is of limited help in determining the variety of moral judgments implicated by moral disagreement and irrelevant influences. Most of it is focused on harm-based transgressions, in particular sacrificial dilemmas. Yet there is more to morality than harm and sacrificial dilemmas (Bauman et al., Reference Bauman, McGraw, Bartels and Warren2014; Graham et al., Reference Graham, Meindl, Beall, Johnson and Zhang2016; Rozin et al., Reference Rozin, Lowery, Imada and Haidt1999). Much more future research will be needed to figure out the extent of disagreement and irrelevant influences in other areas of morality, including judgments about honesty, fairness, loyalty, or authority. Only then can we even hope to know how much of a threat arguments from disagreement and irrelevant influences could pose for moral intuitionism and the methods of moral philosophy.

Even if it turned out that a large variety of moral judgments were untrustworthy, however, another important question remains: What is the proportion of people whose judgments are implicated? Suppose we found that a wide variety of moral judgments (judgments about harm, honesty, fairness, loyalty, etc.) were influenced by framing effects. If this result was due to only 1 percent of the population, the arguments against moral intuitionism and the method of moral philosophy would remain unimpressive. In that case, plausibly, our moral judgments would not have been shown to often be untrustworthy – only the moral judgments of 1 percent of the population. In contrast, if the judgments of 50 percent of the population were influenced by framing effects, this would make for a much stronger case against moral intuitionism and the methods of moral philosophy.

Suppose we had good enough sense of the variety of moral judgments implicated by moral disagreement or irrelevant influences and of the proportion of people responsible for this to come up with an estimate for the proportion of moral judgments that are untrustworthy. How much would be too much? In the context of her meta-analysis of moral framing effects, Demaree-Cotton (Reference Demaree-Cotton2016) claims that 20 percent would not constitute “a large probability of error” (p. 19), and even adds, “I might be happy to accept the possibility that my moral judgments are off-track 20% of the time” (p. 17). This sentiment is echoed by May (Reference May2019, p. 8) and by Sauer (Reference Sauer2018, p. 76).

Others disagree and have argued that 20 percent (or even 10 percent) would be a problem. As one reason, McDonald et al. (Reference McDonald, Yin, Weese and Sinnott-Armstrong2019) point to the importance of morality (Chapters 14, 21, and 22, this volume). Mistakes in moral judgments can lead to hurt feelings, antagonism, bad laws, and even war. Therefore, it is crucial that we get them right. Moreover, McDonald et al. argue that we would not accept scientific judgments if they were off-track 20 percent of the time. Yet the stakes in science are frequently much lower than in morality. This makes it hard to see why we should be happy to accept it if 20 percent of moral judgments could not be trusted.

24.5.4 Does the Research Tell Us about the Right Kind of Moral Judgments?

Another question is whether the studies cited throughout this chapter tell us about the right kind of moral judgments. In particular, some (e.g., Bengson, Reference Bengson2013) have argued that these studies were not set up to capture moral intuitions in particular as opposed to other kinds of moral judgments. If this is right, then, even if the cited studies do show that moral judgments are often not trustworthy, this might not be a problem for moral intuitionism, which is specifically about moral intuitions. However, there still might be a problem if inference is needed to determine that a particular moral judgment is indeed a moral intuition. If so, then moral intuitionists would need to explain how those moral intuitions could be justified independently of inference.

In a similar vein, other have claimed that participants in the kinds of studies that moral psychologists often run tend to give knee-jerk reactions without reflecting deeply, but that the moral judgments whose trustworthiness is relevant to moral theories and the practice of moral philosophy are more considered or reflective moral judgments (e.g., Kauppinen, Reference Kauppinen2007). It is not clear how this objection can be used to defend moral intuitionism, since reflection involves inferences such that the reflective moral judgments will not be justified noninferentially, as moral intuitionists claim. While the objection is more promising when applied to the practice of moral philosophy, various responses to it have been offered (for a sample, see Sytsma & Livengood, Reference Sytsma and Livengood2015). For one thing, it rests on the idea that it is indeed only (or at least mostly) reflective moral judgments that matter for the practice of moral philosophy, which is controversial. Moreover, there is some evidence to suggest that reflective moral judgments may be susceptible to many of the same morally irrelevant influences we mentioned in Section 24.3.3 (Rehren & Sinnott-Armstrong, Reference Rehren and Sinnott-Armstrong2022; Schwitzgebel & Cushman, Reference Schwitzgebel and Cushman2015), which would speak against their trustworthiness.

Some philosophers have offered more conciliatory takes on these objections. Instead of denying that the empirical research can tell us anything of note for the way most moral philosophers currently do their work, they acknowledge that it can pose significant problems, but then argue that there are ways to improve philosophical practice to alleviate them. For example, Tiberius (Reference Tiberius and Kirchin2013) has proposed an empirically informed kind of reflective equilibrium that she suggests may be used to produce philosophically informative moral judgments not subject to the kinds of influences that make unreflective moral judgments untrustworthy. Still, if this kind of reflective equilibrium requires inference, then this move cannot be used to show that any moral judgments are justified noninferentially, as moral intuitionists claim.

24.5.5 The Expertise Defense

Finally, some philosophers claim that they are experts about what is moral (for a survey, see Nado, Reference Nado2014). Proponents of this expertise defense argue that although the empirical research may reveal moral judgments of nonphilosophers to be untrustworthy, judgments of individuals who have had training and practice in (moral) philosophy will be mostly immune to this problem. They are, in effect, expert moral judges. Yet it is the moral judgments of such experts that feature in moral philosophy and meta-ethics. Thus, the method of moral philosophy is not in danger (and maybe neither is moral intuitionism).

One question about this reply is what philosophers’ moral expertise would consist of (Weinberg et al., Reference Weinberg, Gonnerman, Buckner and Alexander2010). Various suggestions have been made. For example, Cameron et al. (Reference Cameron, Payne and Doris2013) have suggested that moral expertise will centrally involve the ability to differentiate the features of one’s emotional experiences. Others have argued that moral expertise can be understood by analogy to linguistic expertise (Driver, Reference Driver2013) or to mastery of skills like driving or playing chess (Dreyfus & Dreyfus, Reference Dreyfus and Dreyfus1991; Ryberg, Reference Ryberg2013).

Whether philosophers are moral experts in a way that will make them impervious to the pull of the arguments from disagreement and irrelevant influences is partly an empirical question. If moral philosophers are indeed experts, then they should not be subject to the types of disagreements and influences revealed by the studies we have surveyed. There is some reason to doubt this. For example, Schwitzgebel and Cushman (Reference Schwitzgebel and Cushman2012) find evidence for order effects among professional moral philosophers comparable to the order effects exhibited by nonphilosophers. Tobia et al. (Reference Tobia, Chapman and Stich2013) found that moral judgments of both philosophers and nonphilosophers were influenced by an unconscious cleanliness prime. Clearly, though, two studies are not enough to draw any strong conclusions about whether moral philosophers are experts. More research will be needed.