1.1. Objections to harm reduction

Politically, most of the opposition to harm reduction in drug policy comes from those who have highly retributive impulses, and so would like to see users punished “as a matter of principle” regardless of whether it serves any constructive purpose (MacCoun, Reference MacCoun2013). Even among those who are reluctant to consider drug consumption malum in se, many regard the fact that it is illegal as sufficient to make users deserving of punishment. To some degree, however, this insistence on punishment is a matter of preference, since the strength of retributive sentiment is subject to considerable variation in the population. In the discursive public sphere, therefore, opponents of harm reduction typically buttress their punitive impulses with arguments intended to expose flaws in the case for harm reduction. There are three important objections to the policy approach that are commonly made.

The first argument, and by far the most important, is the allegation that the two components of the harm reduction approach work at cross-purposes, insofar as the attenuation of harm is likely to increase prevalence through a moral hazard effect. This concern is greatest with paternalistic prohibitions such as drug control, where both the chance of arrest and the probability of adverse health impacts serve as a major deterrent to use. To the extent that an SCS eliminates both of these expected costs, it seems likely to reduce the deterrent effect of the prohibitionist policies in place and so lead to increased incidence at the marginFootnote ⁷ (thus, critics often describe such policies as “enabling” drug abuse). This concern, it should be noted, is exacerbated in many public debates by advocates of harm reduction who treat the prevalence of drug consumption as entirely exogenous to the calculation of costs and benefits (a view that is encouraged by the so-called disease model of addiction), or else favor legalization, and so recommend supervised consumption as an alternative to prohibition. Advocates for either position simply ignore the moral hazard problem, because they do not consider it a problem.

In this discussion, however, I have been adopting a philosophical formulation of the model that features a commitment to both prohibition and attenuation of harm. Theorists who endorse this approach do need to worry about moral hazard problems. The correct way to take these effects into account, however, is not to adopt a zero-tolerance attitude toward moral hazard, but rather to treat these effects as one more factor that must be taken into consideration when determining the appropriate threshold for the transition from prohibition to attenuation. If the attenuation of harm is likely to produce some increased incidence, then one should impose a somewhat higher enforcement threshold, with the recognition that there will be some backsliding over that threshold once the harm reduction policy is implemented. It is also important to recognize that the moral hazard effects are likely to be significant only with paternalistic prohibitions. Because the harms that are targeted by these policies are primarily borne by the individual who is choosing to do x, attenuation of those harms necessarily reduces the disincentive to do x and so can be expected to increase its prevalence. With other-regarding actions, by contrast, attenuating the harms may not generate any moral hazard effect at all, since the harms are borne by third parties. Because the behavior that I am concerned with in this paper primarily involves the imposition of harm on others, the concern about moral hazard is not central.

The second major objection to the harm reduction approach involves concern over the expressive effect of abandoning the commitment to complete prohibition. On the prohibition side of things, we declare a certain action to be intolerable, and yet on the attenuation side of things, we wind up tolerating it. In particular, it is difficult to formulate a harm reduction policy that does not require, in certain instances, that public officials refrain from punishing someone who is observed performing the supposedly prohibited action.Footnote ⁸ This can also generate a moral hazard effect by undermining the deontic force of the prohibition, reducing the stigma associated with the action, or even just by sending mixed messages about its prohibited status. But it is also subject to the more principled objection that harm reduction normalizes the behavior in a way that undermines the prohibitionFootnote ⁹ (again, one can see this with SCSs, which are often portrayed as a Trojan horse for legalization, on the grounds that they may increase social acceptance of hard drug use).

This argument suggests that, while it may be permissible to implement harm reduction policies de facto, they should never be adopted de jure, because of this expressive effect. The point about de facto adoption is important, because there are very few policy domains in which complete elimination of harmful behavior is attainable. In these areas, the limits of prohibition are usually imposed by the budgetary constraints of enforcement agencies. In each major category of crime, for instance, the police do not come anywhere near complete eradication, and so typically spread their efforts over multiple categories (which leads to the commonly voiced complaint, among those arrested for misdemeanors and public order offenses, that there are “unsolved murders” the police should be attending to). The police also, in conjunction with various social welfare agencies, offer support services to victims. What some of these amount to are harm reduction policies in everything but name. And in a sense, it also provides a response to the expressive critique, because it shows that one can adopt a harm reduction policy without necessarily publicizing that fact. But second, it shows that these policies are already common and so should not be so difficult to explain to the public in a way that will avoid mixed messaging. Furthermore, it seems unnecessarily cruel to refuse help to victims, on the grounds that this “sends the wrong message” to potential perpetrators, in cases where one has no realistic expectation of preventing those perpetrators from acting.Footnote ¹⁰

Finally, there is the concern that adoption of a harm reduction approach will lead to premature abandonment of enforcement efforts or missed opportunities for prohibition (MacCoun, Reference MacCoun2013, 92; Weinstock, Reference Weinstock2013, 145). Attenuation of the harms imposed may induce complacency or sap the motivation of various constituencies that might otherwise press for elimination of the behavior. Although conceptually somewhat different from moral hazard, this is also a feedback mechanism that would have the effect of increasing prevalence of the prohibited behavior. Critics of harm reduction often express skepticism about the calculations involved in determining the enforcement threshold, based on the suspicion that more effective prohibition could be achieved through greater commitment or effort. Part of the popularity of “tough on crime” policies, for instance, is that the members of the general public tend not to see (or suffer) the cost of enforcement efforts, whereas the effects of crime are a great deal more apparent, which inclines them to believe that greater prohibition could always be achieved through tougher enforcement. Not only does attenuation of harm contribute to this perception of laxness, there is also the perception that it papers over the problem, by increasing social acceptance of the failure to achieve proper enforcement.

Again, it is important to acknowledge that this is a reasonable concern, but that it does not constitute an insuperable obstacle to the implementation of harm reduction policies. What it suggests, instead, is that the development and implementation of such policies must be grounded in high-quality social science, with real metrics and measurable data, and not just ad hoc or intuitive perceptions of social impacts. It also implies, however, that while public input will always be a crucial element in the development of such policies, decision-making with respect to their implementation generally should not be made by the broader public (for example, the American practice of having democratic elections to appoint key players in the criminal justice system—sheriffs, district attorneys, and even judges—is widely blamed for the overly punitive character of that system, because the public considers the elimination of crime to be an attainable objective and so favors a zero tolerance approach [Pfaff, Reference Pfaff2017]). While implementation should always be subject to democratic control through high-level leadership, the specific formulation of harm reduction policies, and in particular, the specification of the recommended enforcement threshold, must necessarily be informed by expert judgment.

2.1. Groupishness

Studies of human infants have shown that we impose a very coarse classificatory system on the world prior to the development of linguistic competence (and thus presumably with little or no cultural learning). Very young infants exhibit a complex set of responsive dispositions toward human faces, in particular, which reveal classificatory tendencies (Marquis & Sugden, Reference Marquis and Sugden2019). There is a widely accepted line of argument, however, which suggests that it is unlikely that humans have any innate disposition to classify individuals on the basis of race, or even notable features associated with ancestry, simply because these group differences were not present in the environment of evolutionary adaptation (Cosmides et al., Reference Cosmides, Tooby and Kurzban2003). For over 90% of human history, people lived in small, culturally homogeneous groups that seldom exceeded 100 or so individuals. Furthermore, while most groups appear to have been exogamous, marital exchange typically occurred with neighbors who were geographically quite proximate. As a result, human infants grew up in social environments in which they were not exposed to significant racial differences, suggesting that there would have been no adaptive value to engaging in such classification.

This is the good news, as far as the eliminativist project is concerned. The bad news is that the small group environment in which humans evolved was also quite competitive, involving high levels of solidarity among insiders combined with distrust and antagonism toward outsiders. This appears to have led to the evolution of a set of cognitive and behavioral heuristics that encourage a high level of “groupishness” (Haidt, Reference Haidt2012; Richerson & Boyd, Reference Richerson, Boyd and Nesse2001). This involves, first, a propensity to classify individuals in the social environment into categories of “in-group” and “out-group” (or “us” and “them”); second, a disposition to act more cooperatively toward and to experience more sympathy for in-group members; and finally, a disposition to act less cooperatively, or more aggressively, toward out-group members. The cognitive and behavioral aspects of this psychological complex tend to come as a package. As Henri Tajfel and John Turner have observed, in a survey of the literature, “the mere perception of belonging to two distinct groups—that is, social categorization per se—is sufficient to trigger intergroup discrimination favoring the in-group” (Reference Tajfel, Turner, Hatch and Schultz2004, 56). They do not hesitate to describe this as the psychological basis of ethnocentrism, or what we might call exophobia (i.e., social difference aversion).

Apart from its cultural universality, the fact that such a complex suite of behaviors can be activated so reliably and easily, even in fairly young children, suggests that the basic psychological complex is not the product of cultural learning or socialization. Furthermore, once group identification has been triggered, it tends to override other potential sources of solidarity (in one fairly typical study, an explicitly random classification of individuals into groups proved to be “a more potent determinant of discrimination than perceived interpersonal similarities and dissimilarities not associated with categorization into groups” [Tajfel & Turner, Reference Tajfel, Turner, Hatch and Schultz2004, 57]). These two factors combined might suggest that exophobia is an inevitable feature of the human condition, and indeed, surveying the social psychology literature on this subject can be quite discouraging. There is, however, one bright spot, which is that our evolved psychological complex appears to be agnostic about the basis of the classificatory scheme that triggers in-group identification. Furthermore, once a particular basis has been imposed (or becomes focal), it tends to occlude all others. As a result, our groupishness can be manipulated or spoofed (in some cases, easily so) (Voorspoels et al., Reference Voorspoels, Bartlema and Vanpaemel2014; also Brewer & Miller, Reference Brewer, Miller, Miller and Brewer1984). This appears to provide the psychological grounds for an early and ubiquitous cultural adaptation, which is that all known small-scale societies adopt symbolic markers, which serve to provide easily identifiable visual cues of group identity (Boyd & Richerson, Reference Boyd and Richerson2005, 211–212). These are entirely conventional and so must be learned by children or initiates to the group.

Because of this, there is a very narrow sense in which it is correct to say that racism is learned. Since there is nothing inevitable about the use of physical traits linked to ancestry as the basis of in-group classification, American children are not born knowing that skin color, rather than hairstyle or clothing type, is the central characteristic demarcating “us” from “them” in their society. But of course once they pick up on this—and there is a wide range of different forms of social behavior that can convey this information, many of which involve no intergroup antagonism—then everything else follows quite automatically. Most importantly, no one must teach them to experience greater sympathy and concern toward in-group members, or to behave aggressively and non-cooperatively toward out-group members. In other words, no one teaches us to be exophobic, or endophilic; cultural patterns merely provide direction to these impulses.

Given this understanding of our exophobic dispositions, it is not difficult to see how certain efforts to eliminate racism can become counterproductive. The most direct strategy to reduce racial animosity is to reduce the salience of that particular classification in the social environment, so that individuals refrain from drawing the in-group/out-group distinction along racial lines (e.g., it is preferable for residents of Chicago to see themselves up as “White Sox fans” vs. “Cubs fans,” rather than “white” vs. “Black”). This approach, however, runs contrary to the practice, which has recently become popular in the U.S., of encouraging heightened “race consciousness” (along with firm rejection of efforts to practice “color-blindness”).Footnote ¹⁵ The suggestion is that individuals should be encouraged to continue seeing racial difference and to categorize both themselves and others on this basis, and yet should suppress the exophobic behavioral complex that is associated with it.Footnote ¹⁶ The standard argument for this position is that certain forms of racial inequality cannot be combatted without drawing attention to them, or counteracting them in a targeted way, and so a strategy that distracts from racial division will have a tendency to perpetuate inequality. If this is true, however, it is not difficult to see how it could generate a paradoxical situation, in which it is impossible to advance the struggle against racism without increased activation of the psychological dispositions that are responsible for it. Many individuals will succeed either in suppressing this or channeling it in a productive direction, but many others will not. The fact that categorization triggers discrimination, therefore, imposes intrinsic limits on the eliminativist program.

2.2. Authoritarianism

One of the striking features of authoritarian political movements is that the complex of attitudes they channel is so consistent across different places and times. They typically bundle together a set of superficially disparate concerns, including support for strong, decisive leadership, usually exercised by a single man; adherence to conventional morality and social norms; highly punitive attitudes toward social deviance, with a preference for forceful response; and hostility toward immigrants and outsiders.Footnote ¹⁷ This has generated the suspicion that some relatively unified, underlying personality trait may predispose individuals to exhibit this complex of attitudes.Footnote ¹⁸ Serious efforts to characterize it were undertaken in the wake of the Second World War, motivated by the effort to understand the psychological underpinnings of support for fascism (culminating in the publication of The Authoritarian Personality, by Theodor Adorno, Else Frenkel-Brunswik, Daniel Levinson, and Nevitt Sanford in 1950).

It is important to emphasize at the outset that the empirical research driving these discussions is not as uncontroversial as that which informs our understanding of groupishness and exophobia. The latter phenomenon is easily elicited both in naturalistic settings and laboratory conditions, and the core findings have been extensively replicated. Personality studies, on the other hand, make extensive use of questionnaires to elicit attitudes, then use statistical methods to identify connections between these different attitudes. The correlations are usually not enormously strong, data are often open to interpretation, and there are concerns about both the validity of questionnaire responses and their connection to behavior. Early efforts, including The Authoritarian Personality, also suffered a number of specific methodological flaws that were widely criticized. Modern work, such as Stenner’s The Authoritarian Dynamic (Stenner, Reference Stenner2012), strives to avoid these pitfalls, but it remains wedded to the basic methodological precepts of personality psychology.

With these caveats in place, one may turn to the central finding, which is that certain individuals possess a set of reactive dispositions, distinct from groupishness, that generates support for authoritarianism. More precisely, what they possess seems to be an exaggerated form of the psychological complex we all possess that forms the basis of our “norm psychology” (Sripada & Stich, Reference Sripada, Stich and Carruthers2007). This features two intrinsic motives that are activated by the observation of rule-following behavior, first a desire to conform, and second a desire to punish deviance (Sripada & Stich, Reference Sripada, Stich and Carruthers2007, 287). We can be fairly confident that these dispositions are innate, not learned, because they form part of the psychological substratum that supports the development of cultural transmission (Boyd & Richerson, Reference Boyd and Richerson2005, 83–97). But if they are under biological control, it stands to reason that there will be variability in their strength across individuals. This suggests that some people will be “hyper-conformist” in their basic orientation: not only highly motivated to conform in their own behavior, but highly distressed by a lack of behavioral uniformity among others, and angry and punitive in their reactions. This gives them the set of reactive emotional dispositions that we associate with “intolerance,” which they may or may not succeed (or even try to succeed) in controlling. For reasons that are less well understood, this sort of intolerance also appears to generate support for “strong leadership” on the conventional model (Stenner, Reference Stenner2012, 19).

One other important and widespread finding is that these dispositions are not always evident in everyday social interaction, but are triggered by particular social conditions that are experienced as threatening (Doty et al., Reference Doty, Peterson, Winter, Jost and Sidanius2004, 86–108). Such threats include non-conformity or lack of uniformity, either exhibited with respect to normed behavior and identity, or expressed with respect to belief and values. Lack of confidence in leadership also appears to trigger the same reaction. The major consequence of this body of research has been to suggest an exception to the basic findings of contact theory, or the view that under appropriate conditions face-to-face interaction between individuals belonging to different racial or ethnic groups will tend to reduce prejudice (Allport, Reference Allport1954; Pettigrew, Reference Pettigrew1998; Pettigrew & Tropp, Reference Pettigrew and Tropp2006). While there is considerable evidential support for this hypothesis, there are also important exceptions, as well as a more general concern that positive results may reflect selection bias, as people who are relatively open to difference put themselves in a position where they are more likely to experience such contact. In some cases, exposure to difference appears to generate precisely the sense of threat that triggers the authoritarian response.

Some have drawn from this the alarming conclusion that liberal-democratic societies are self-undermining, because they encourage various forms of diversity that in turn give rise to authoritarian backlash, a claim that has been met with some criticism.Footnote ¹⁹ Yet, one need not endorse this pessimistic view to see that there are nevertheless cautionary lessons to be learned. First, it suggests that certain attempts to eliminate racism, especially those involving increased contact, may become counterproductive over time. While they may initially enjoy some success, as they experience uptake from individuals who are relatively open to experience, when extended to the broader population, they may begin to provoke authoritarian backlash. This seems particularly likely with initiatives that seek to provoke “uncomfortable” conversations and encounters, based on the theory that this is necessary in order to overcome majority self-deception about racial privilege (Sue, Reference Sue2015). Second, even in cases where anti-racism initiatives do not become directly counterproductive, they may still suffer from a “last 10 per cent” problem dealing with individuals who are highly susceptible to normative threat.

2.3. Stereotyping

The concept of a “racial stereotype” was introduced in order to describe false beliefs held by individuals about other racial groups, acquired through culturally transmitted representations (such as media portrayals).Footnote ²⁰ A major virtue of intergroup contact was that it was expected to cure individuals of these stereotypes, by giving them direct exposure to members of other groups and thus an opportunity to learn the truth. Stereotyping is taken to be a major source of bias in interaction between members of different racial groups. Furthermore, while individuals who are overtly hostile to other racial groups will typically exhibit bias in their interactions with them and will be aware of that bias, stereotyping can lead individuals to exhibit bias without any awareness that their judgment is being influenced in this way. There has been an explosion of interest in such forms of “implicit bias,” particularly with the discovery that many seem linked to subconscious associations that individuals hold (which can be elicited through an implicit association test).Footnote ²¹

Over time, however, the term stereotype came to be used in a wider sense, to refer not just to false beliefs about others, but rather to any sort of generalization about groups. These generalizations are typically false when interpreted universally, since they would seldom be true of each single individual in the group, but are often true when interpreted statistically, as claims about averages or probabilities. A stereotype that is true in this sense can be referred to as accurate (Ryan, Reference Ryan2002). For example, the claim that “Asians are studious” would typically be considered a racial stereotype in America, even though it is accurate (on average, Asian students in U.S. elementary and secondary schools spend more time doing schoolwork than members of the other major census category groups) (Hsin & Xie, Reference Hsin and Xie2010). The harm of stereotyping occurs when people make assumptions about a specific individual based on qualities ascribed to the racial group to which that person belongs, and are either slow or unwilling to update their priors based on evidence derived from observation of the individual.

Confusingly, stereotypes are often described as a source of bias, regardless of whether they are accurate or inaccurate, even though the term bias suggests a deviation from statistically valid inference (Jussin, Reference Jussin2012). Many discussions of implicit bias, for example, point to the association that many people have between Black men and violent crime in the U.S., or between Indigenous youth and substance abuse in Canada. Both of these associations, however, are based on accurate statistical generalizations about these groups (Morenoff, Reference Morenoff, Ruter and Tienda2005; Sikorski et al., Reference Sikorski, Leatherdate and Cooke2019). As a result, these stereotypes have proven quite resistant to eradication, precisely because they tend to be reinforced by real-world experience. Denouncing such beliefs as racist, or as biased, as many educators and some philosophers (e.g., Basu, Reference Basu2019) have been inclined to do, sets up a conflict between epistemic and moral values that is unlikely to be definitively resolved in favor of morality.

When the term “racism” is used in this expansive sense, to include epistemically valid beliefs and inferences, then it is unlikely ever to be eliminated. And yet even restricting one’s attention to the epistemically invalid generalizations that people are inclined to make, which fit more closely the traditional definition of “racist,” it is not clear that they can ever be suppressed entirely in the population at large. For example, when dealing with a stranger, many would consider it epistemically permissible to use that person’s racial group membership to establish a baseline probability (or a Bayesian prior) with respect to certain characteristics, with the expectation that this would be updated to reflect any new evidence that comes along. What we are not permitted to do is assume that the characteristic must be true of the individual (this would involve essentializing group differences in a way that is widely acknowledged to be racist), or to resist updating in the light of evidence. And yet these rather subtle distinctions are unlikely to be grasped by a substantial percentage of the population, much less scrupulously adhered to. As a result, it is difficult to imagine a situation in which people continue to draw the relevant group distinctions and yet refrain from overgeneralizing on their basis.

Given the centrality of frequency-detection and expectation-generation to ordinary human cognition, it is also exceedingly unlikely that the central mechanism that gives rise to stereotyping could ever be suppressed (Monteith et al., Reference Monteith, Sherman and Devine1998). Thus, the only real way to prevent individuals from reasoning on the basis of racial stereotypes is, again, to reduce the salience of racial cues in the environment, so that individuals refrain from categorizing the regularities being observed in the environment along racial lines. As noted earlier though, this is in tension with the practice of most current anti-racism efforts, which have been sharply critical of the attempt to cultivate color-blindness in everyday social interaction. At the very least, there is a conflict between this desire to cultivate race-consciousness and the effort to deter individuals from reasoning in accordance with racial stereotypes.

2.4 The case for pessimism

The best-case scenario for the elimination of attitudinal racism would be if these habits of thought and action constituted entirely learned behavior, in which case they could be discontinued simply through a concerted effort to stop teaching them. Even here, of course, one must be cautious. There is a good case to be made that religious beliefs are also taught, and yet several attempts to eliminate them have proven unsuccessful—to the point where Enlightenment liberals have learned to accommodate themselves to their persistence. As I have tried to show in the previous three subsections, racism is closely tied to a set of cognitive heuristics that are widely thought to be innate. Of course, one must be cautious in the conclusions that one draws from this, because innate dispositions do not inevitably find expression in behavior—we can learn to suppress them, and there is no doubt that many people living in modern, pluralistic societies do just that. The problem in the case of racism is that these heuristics seem inextricably tied to other, legitimate forms of social cognition, making it difficult to police their employment even among people of good will. Most of us like the feeling of belonging to a team, club, group, or school of thought; we appreciate the sense of ease afforded by life in a close-knit community where behavioral expectations are settled and uniform; and we rely on social categorization and expectations derived from past experience to manage everyday interactions, especially with strangers. It is very difficult to draw a sharp line between harmless and harmful exercises of these dispositions.

The situation is somewhat exacerbated by the tendency, in recent years, toward adoption of increasingly expansive conceptions of racism (Blum, Reference Blum2002; Carmines et al., Reference Carmines, Sniderman and Easter2011; Sniderman et al., Reference Sniderman, Piazza, Tetlock and Kendrick1991). Moving the goalposts makes it practically inevitable that racism will resist efforts at eradication. To pick just one example, the recent focus on the elimination of racial “microaggressions” makes the elimination of racism tantamount to the abolition of rudeness in interpersonal relations involving individuals of different races, which I think would strike most people as an unobtainable outcome (Sue, Reference Sue2010). And yet even if one sticks to the old-fashioned conception of racism, focused on explicitly abusive speech, prejudice, and discrimination, it seems unlikely that racism will ever be completely eliminated. My argument in support of this contention is twofold.

First, there is the fact that many of the strategies that have been employed to combat racism have provoked backlash, making them directly counterproductive. For example, with respect to the elimination of racial slurs, the development of a strong taboo against certain uses of language can have the perverse effect of transforming these terms into swear words. This is sometimes referred to as the “paradox of profanity,” where the effort to suppress particular terms winds up promoting their use, precisely because of the censorious reaction that their use generates (Bergen, Reference Bergen2016, 216). One can see this at work in the effort among Christians to stop people from taking the lord’s name in vain (in expressions such as “God damn it” or “crisse de tabarnak”). The fact that people were offended by these sacrilegious expressions is part of what made it satisfying to say them. Accordingly, as their use became increasingly common and people became less offended by them, these terms became less satisfying, as a result of which people began to shift toward words that retained the power to shock. Because of this, a more measured response, aimed at discouraging the use of certain words, rather than prohibiting them altogether, may wind up doing a better job at reducing their use in the long term.

A similar concern arises with the attempt to heighten race-consciousness, which is often seen as essential to the eradication of racial injustice, yet seems unlikely to provoke purely pro-social manifestations of groupish thinking. In particular, the attempt to persuade majority group members, such as whites in the U.S., to adopt an explicitly racial identity seems likely to become counterproductive. This is particularly important in the field of “whiteness studies,” where many assert that white privilege is based, at least in part, on the fact that whiteness is an unmarked racial identity (Kolchin, Reference Kolchin2009). And yet it is not clear that the proposed solution, which involves cultivating an explicitly racial white identity in the majority population, is going to produce greater generosity and tolerance in relations with minorities. It seems also to be provoking a resurgence of traditional white nationalism. Even if the intervention produced one white nationalist for every ten “repenters” (Bright, Reference Bright2023) it clearly would not be achieving its goal.

Second, there are reasons to worry that the last 10% problem will be quite severe in the case of racism. If one imagines racial intolerance as subject to a normal distribution, the problem is that there are relatively few interventions that specifically target the tail end, where the most problematic cases lie. In part this is due to the relative success of anti-racism initiatives, which despite falling short of eliminating the problem entirely, have nevertheless succeeded in making the overt expression of racist views rather uncommon in formal settings. As a result, despite the fact that most anti-racism initiatives are not overtly punitive, those who harbor racial animus are generally loathe to self-identify, which in turn makes it difficult to focus remedial efforts on them. The majority of anti-racism initiatives seem to be focused, therefore, on further displacing the mean, presumably in the hope of reducing the number of individuals falling below a certain threshold of tolerance. Attempting to shift the entire population is obviously a very inefficient way to deal with the problems posed by outlying cases. Apart from being costly, it also runs the risk of becoming counterproductive, as individuals close to the mean are obliged to redo exercises whose point they have already fully internalized.

It is also worth noting that the commitment to freedom of expression puts rather severe constraints on the willingness of the state to employ coercive power to limit racist speech. The zero-tolerance orientation has, therefore, generated significant pressure to reduce traditional inhibitions against legal regulation of speech and has led many institutions to adopt internal norms to punish racist speech. Some U.S. universities, for example, created “bias incident” hotlines and incident reporting systems, so that episodes of racist speech could be logged and possibly subjected to disciplinary action. Much of this subsequently descended into chaos, first with the eruption of conflict over Israel/Palestine and the emergence of intractable disagreement over what constituted “hate speech” in this domain, and then with the interventions of the Trump administration. But even if the efforts had been successful, the attempt to impose a punitive regime on speech in face-to-face interaction had the effect of driving a great deal of social deviance online, where subcultures that tolerate or promote racist speech have been thriving. The overall result is that the enforcement threshold has been, and is likely to remain, relatively low for all but the most targeted and offensive racist speech. Thus, the “fact of unenforceability” is likely to prevail with respect to the general prohibition on attitudinal racism.

3.1. Tolerated self-isolation

There is a long-standing tension in liberal societies between the commitment to individual freedom, which, when combined with background pluralism, generates sorting of the population along various dimensions of difference, and the goal of promoting integration in various institutions, which is applauded on the expectation that it will produce greater tolerance and mutual understanding. Reasonably large constituencies favor policies aimed at achieving diversity (or “representation”) in major social institutions, as well as with respect to work and living arrangements (Anderson, Reference Anderson2010). The usual objection to this essentially teleological objective is that it may interfere with individual liberty or freedom of association, prohibiting or penalizing individuals for selecting social environments that match their preferences. This has set the stage for what often seems like a contest between incompatible values or political commitments.

There is, however, an important ambiguity in the commitment to integration, which is that the goal itself provides no specification of the level at which integration must be achieved (Soon, Reference Soon2023). For example, in the U.S., the concern over residential integration has typically focused on the racial composition of neighborhoods (primarily because of educational funding and access, which is local). In Singapore, by contrast, concern over racial integration extends beyond neighborhoods to encompass individual apartment buildings, each of which must respect the quotas that have been placed on the four recognized racial groups (Heath, Reference Heath2023, 288–292). The latter policy obviously entails much greater restrictions on sorting, and thus more active interference with individual liberty, than the former. One way to minimize the interference is, therefore, to formulate the integration objectives at the highest level compatible with the background rationale for the policy.

There has been a trend in the U.S., however, to move in the opposite direction, imposing diversity requirements at increasingly lower levels. One can see this most clearly in popular culture, especially television and movies. Whereas it was once considered an adequate achievement for television shows or movies in general to feature adequate racial diversity, this has become increasingly supplanted by the expectation that each television show, or every single movie, be internally diverse.Footnote ²⁴ The concern is presumably that if racial diversity exists only within the category, racist individuals can still avoid exposure to diversity by making careful within-category selections, something that can be prevented only by introducing diversity into every single product in the category. In this case, however, the motivation for the elimination of sorting seems to be moral condemnation of the preference that gives rise to it, rather than any specific harm that it generates for others.Footnote ²⁵ Thus, a harm reduction approach would be willing to countenance a higher degree of self-sorting, treating majority endophilia roughly in the same way that minority endophilia is treated today—permissible in the domain of private, primarily self-regarding choice, so long as it does not generate demonstrable harm to others.

3.2. Non-amplification

During the 20th century, access to public discourse was controlled by a number of very powerful gatekeepers—primarily the editors employed by major newspapers, periodicals, publishing houses, and television stations. Once a general social taboo emerged against the expression of racist attitudes in the 1970s, the result was a public sphere in which major communication channels were essentially free of overtly racist speech. This state of affairs changed dramatically with the development of the internet and social media, which had the effect of removing the gatekeepers, allowing private individuals to achieve unmediated access to potentially enormous audiences. The result was a certain measure of shock in many quarters at the level of racist speech that had traditionally been suppressed but was now finding its way into the public sphere.

The initial impulse of many has been to demand censorship of this speech, and indeed most private social media platforms have been persuaded to adopt content moderation policies that impose limits on the extremes of anti-social online behavior. But because of the commitment to free speech these policies stop short of re-establishing the older norms of decorum. Without getting into the details of these debates, one may observe that certain interventions aimed at curtailing racist speech can have the unwanted effect of amplifying it. For example, the practice of “calling out” offensive speech, as a way of expressing one’s moral condemnation, also has the effect of drawing attention to it.Footnote ²⁶ This is often accompanied by an attempt to recruit others to the effort, which can result in the original, offensive message being reproduced thousands, hundreds of thousands, or in some cases millions of times. The assumption underlying these efforts would appear to be that the punitive effect of the mass call-out will serve as a deterrent to others. And yet it is not clear that the broadcast effect will be outweighed by the deterrent effect, in such a way as to result in a net decrease in the incidence of racist speech online. The attention is itself is often rewarding to the offender, since a great deal of racist speech is rather self-evidently attention-seeking, particularly among the young. There is also a reasonable concern that the amplification may make life worse for minorities, by causing them to overestimate how common these offensive views are in the population (Hughes, Reference Hughes2024, 92).

One can see a similar issue with the emergence of far-right political parties, which typically cater to a combination of authoritarian and exophobic impulses. Most are aware that the expression of overtly racist views will alienate more voters than it will attract, and so maintain relatively strict message discipline in their public pronouncements. Because of this, their political opponents usually seize upon anything said that can be construed as racist and draw attention to it, in an effort to expose the true nature of these parties. In doing so, however, they also provide significant amplification of those attitudes. The assumption seems to be, again, that the more widely the remarks are denounced, the more effectively the attitudes will be punished, or at least silenced. Empirically, however, one might wonder whether this is so. Most obviously, it gives these parties the capacity to hijack public deliberation on practically any subject, simply by saying something inflammatory, which, in turn, shifts attention to their preferred issues.

3.3. Sublimation

Sigmund Freud believed that the various forms of instinctual renunciation required for civilization were achieved not only through straightforward repression but also through sublimation—which is to say, through the provision of substitute gratifications (Freud, Reference Freud2010). This is contrary to the dominant impulse of the moralist, whose inclination is always for straightforward repression of anti-social instincts. With respect to the human taste for violence, for example, many would like to see it overcome completely, and so the prevailing compromise, in which the impulse is redirected in ways that make its expression less harmful (e.g., toward violent video games and movies), is seen as distinctly second best. Indeed, the accusation is often made that these cultural products have the opposite effect, of stimulating violent impulses that would otherwise be absent or dormant, which then risks increasing the level of genuine violence.

Again, these are complex debates that I am not in a position to adjudicate. I would like to observe, however, that the impulse toward groupishness, which appears to be central to a great deal of racial conflict, is also susceptible to sublimation through the cultivation of other particularistic loyalties. It is perhaps noteworthy, in this context, that the two most well-integrated institutions in American public life, namely, the military and professional sports, involve a social context that provides very strong group identity along non-racial lines. These institutions are distinctive in that they provide racially cross-cutting allegiances. Similarly, there is evidence that individuals in the U.S. who actively identify as gay, lesbian, or bisexual have an “increased propensity for interraciality in a range of interpersonal relationships” compared to heterosexuals (Horowitz & Gomez, Reference Horowitz and Gomez2018). The suggestion is that particularistic group identities crowd one another out, and so the way to combat groupishness is not to suppress it, but rather to redirect it toward less socially pernicious axes of identification (Gaertner & Dovido, Reference Gaertner and Dovido2000).

It is important also not to overlook the occasional contribution of institutions that provide higher level sources of common identity. While cosmopolitan institutions have been relatively ineffective at disrupting racial polarization, special note must be taken of the nation state and of nationalism, which possesses an affective resonance for many people that suggests satisfaction of groupish impulses. Again, just as in the case of violent entertainment, there is the suspicion that these substitute gratifications are not actually preventing the harmful expression of the attitudes being targeted, but that they may, perversely, be encouraging them. The cultivation of “tribal” loyalties, in areas such as fandom or patriotism, is sometimes seen in the same way, as offering a potentially counterproductive form of indulgence to these less-than-universalistic sentiments. A harm reduction approach, by contrast, which accepts the inevitability of such sentiments, is more likely to see these institutional strategies in a positive light, as offering a relatively harmless outlet for what are nevertheless acknowledged to be anti-social impulses.

3.4. Identity-blindness

There has been a trend in recent social justice scholarship toward increasingly vehement condemnation of so-called color-blind approaches to addressing racial inequality. During the early period of the U.S. civil rights movement, when explicit racial discrimination was rampant, it seemed relatively straightforward to assume that eliminating this sort of discrimination—so that individuals would no longer be treated differently depending on their race—would constitute an important form of moral progress. At the same time, many expressed concern that the merely negative action of eliminating barriers to advancement would not be sufficient to bring about equality, and that other, more positive measures (i.e., “affirmative” action) would be required. In more recent years, however, it has been commonly asserted that the merely negative actions typically undertaken under the mantle of color-blindness were active impediments to racial progress, or worse, that they constituted a form of racist ideology. Rather than calling for supplementation of traditional non-discrimination measures, many have called for their abolition (Hughes, Reference Hughes2024, 26–32).

These accusations against color-blind policies are often made amidst a confusing miasma of empirical and normative definitional claims. For example, the commitment to color-blindness is often identified with the empirical belief that racial inequality has been eliminated from American society, and so nothing further need be done to rectify the problem (Neville et al., Reference Neville, Awad, Brooks, Flores and Bluemel2013, 457). Similarly, the commitment to color-blindness is sometimes equated with the view that procedures need only be facially neutral, such that disparate racial impact can be ignored entirely. These uncharitable definitions have the effect of obscuring what is, in fact, an important normative debate over the type of policies that should, or may permissibly, be adopted to combat racial discrimination.

An important recent meta-analysis of the effects of “diversity ideologies” has offered a useful reframing of the issue by drawing a distinction between “identity-blind” and “identity-conscious” approaches (Leslie et al., Reference Leslie, Bono, Kim and Beaver2020). The authors found that, while identity-conscious “multiculturalist” approaches had the most significant effects, a combination of color-blind and meritocratic identity-blind approaches was also effective at reducing stereotyping, prejudice, and discrimination. This generates the suspicion that a great deal of the hostility toward identity-blind approaches stems from the fact that they stop short of rooting out the scourge of race prejudice; they merely prevent or prohibit individuals from acting on whatever attitudes they may have. A “meritocratic” approach, for example, which favors a rigorous emphasis on performance as a basis of advancement to the exclusion of other characteristics, reduces discrimination, but does so without requiring anyone to develop warm feelings toward others, much less affirm anyone else’s identity. For this reason, it can easily be characterized as second best. And yet, if one comes to think that universal affirmation is unobtainable, this second-best outcome may begin to seem attractive.

In this context, it is important to keep in mind the extent to which institutions and policies in pluralistic societies have been improved through color-blindness. For example, the long-standing grievance of African-Americans, of not being able to hail a taxi in major cities, has disappeared from public discourse, largely due to the effects of ride-sharing apps that do not disclose the personal characteristics of riders. Similarly, there is no movement in the U.S. to restore the practice of including a photograph of job applicants on CVs, and countries like France, in which the practice remains prevalent, could probably benefit a great deal from its abolition. There are also simple steps that could be taken in residential housing and rental markets to make racial discrimination practically impossible. All of these policies would have the effect of reducing the harm produced by residual racist attitudes, by making it difficult for those who harbor such attitudes to act upon them.

3.5. Psychological resilience of victims

A common feature of minority group members who have succeeded in hostile or discriminatory environments is the psychological resilience they have exhibited in the face of adversity. In most cases, there are institutions within minority communities that bolster these efforts. In recent years, however, concerns have been raised about the possibility that certain anti-racism efforts within mainstream institutions may be inadvertently undermining the psychological resiliency that minorities require in order to succeed under adversity (Campbell & Manning, Reference Campbell and Manning2018; Lukianoff & Haidt, Reference Lukianoff and Haidt2018). Two factors have been identified that may be contributing to this effect. First, in social environments where content-based restrictions on free speech are prohibited, it has become common to argue that racially abusive speech should be prohibited on the grounds that it is directly harmful, or that it threatens the “safety” of minority group members. Although the initial motivation for such claims was presumably to call for restrictions on speech while formally maintaining adherence to traditional free speech norms, the practice of claiming that speech produces psychological harm appears to have become self-fulfilling in certain cases. Second, the zero-tolerance approach to racism has encouraged individuals to call for third-party intervention in response to even very minor incidents (Miller et al., Reference Miller, Guida, Smith, Ferguson and Medina2018). The result has been a certain tendency toward catastrophization, as individuals are encouraged to claim psychological injury and seek redress in response to minor slights. The result, as Greg Lukianoff and Jonathan Haidt (Reference Lukianoff and Haidt2018) allege, is a type of “reverse cognitive-behavioural therapy,” wherein our moral commitments encourage individuals to act in a way that is diametrically opposed to the habits that encourage psychological resilience.

Again, it is important to be clear about the scope of this claim. I am not recommending that minorities develop psychological resilience in the face of actual discrimination or threats. The issue concerns primarily the response to the expression of negative attitudes, up to and including the use of racial slurs. Advocating psychological resilience in the face of such behavior is fraught, because of the moral judgment that the incidents that require it are contranormative and so should not be happening at all. But once it is accepted that they will never be eliminated entirely, it becomes important to minimize the harm that they do. One way of reducing that harm is to minimize their psychological effects on victims, for example, to ensure that they do not cause significant psychological damage or trauma. This seems particularly achievable with respect to the category of offenses currently described as microaggressions, many of which constitute facially neutral remarks that are targeted for prohibition because of the alleged psychological harm that they cause (Rini, Reference Rini2020; Sue, Reference Sue2010). In such cases, it seems clear that the imperatives of harm reduction may impose burdens of adjustment on both speaker and listener.