THE HARM-BASED ACCOUNT

The most sophisticated version of the harm-based account is offered by Kasper Lippert-Rasmussen. It holds that:

There are three dimensions of this definition that warrant elaboration.Footnote ⁷ First, we should distinguish between two types of harm: the harm inflicted on the recipient of the discrimination, and the harm inflicted on others. If a person harmlessly discriminates against another (we defend this possibility below) but harms a third party, it is open to the harm-based account to hold that the act is wrongful because of the harm to the third party. Would this imply that the person who is wronged is the third party rather than the recipient of discrimination, and also that the third party is a victim of wrongful discrimination, regardless of how the collateral harm results? If so, we doubt the plausibility of these implications, but we leave this issue open, since our objections do not require resolving it.

Second, the harm-based account is most plausibly understood as providing an account of the prima facie or presumptive wrongness of discrimination.Footnote ⁸ If a discriminatory act is harmful, this generates a presumption in favor of that act's being wrong. In some cases, a presumptively wrong act turns out on reflection not to be wrong in any way. This can be seen in some cases of affirmative action, where, since certain well-off individuals have already benefited from unfair background conditions, it is in no way wrong to discriminate harmfully against them in order to remove this unfair advantage. In other cases, a presumptively wrong act of discrimination is pro tanto wrong but not wrong all things considered. In these cases, the pro tanto wrongness is outweighed by countervailing considerations. One example would be discriminating against a member of an unfairly advantaged group who has not been individually benefited. If a male candidate who has not personally benefited from the advantages enjoyed by males is disadvantaged in the implementation of an affirmative-action policy, he is treated pro tanto unfairly. But this unfairness may be overridden by the instrumental benefits of the policy if it achieves a fairer distribution of opportunities between men and women overall.

Third, in order for the full range of implications of the harm-based account to be understood, it must make reference to a complete theory of harm. Such a theory must answer at least two questions: What is the currency of harm? And how is harm to be measured? The currency is the good or goods to which a setback constitutes harm. Specifying the currency helps to characterize harm, as it is the value of the currency that explains why its frustration is bad. For example, if preferences are the currency of harm, the value of preference-satisfaction explains why harm is bad, for harm consists in frustrating a person's preferences. The measure of harm is the scale on which degrees of harm are marked.Footnote ⁹ For instance, should harm be measured counterfactually, by reference to an alternative state that the victim might now have been in had some other sequence of events occurred? And if so, how do we pick out the relevant counterfactuals, and what is the appropriate baseline with which to compare them?Footnote ¹⁰ We return to some of these questions below.

We begin this paper with three questions that are central to an account of the wrongness of discrimination: What are the necessary and sufficient conditions of wrongful discrimination? What explains why discrimination is wrongful? And what factors aggravate the wrongness of discrimination? We can call these the “conditions question,” the “explanation question,” and the “degree question,” respectively. The most attractive version of the harm-based account answers these questions as follows. First, harm is a necessary (though perhaps not sufficient) condition of wrongness. Second, harm is the wrong-making property. That is, discrimination is wrong, when it is, because it is harmful, not simply because harm always accompanies the actual wrong-making property. Finally, the degree of harm done is an aggravating factor. The addition of other aggravating factors, such as the preexisting level of well-being of the victim or membership of a historically disadvantaged group, is also possible. This means that the harm-based account can be non-consequence-focused with respect to the degree question. Again, since our objections do not require rejecting these possibilities, we leave them open.

THE CONDITIONS OBJECTION

The harm-based account has implausible implications in cases of harmless discrimination. We proceed by analyzing a variety of cases in which a discriminatory act seems wrongful even though it causes the victim no harm.

Consider the following case, which is presented by Lippert-Rasmussen:

The harm-based account seems to imply that since the victim is not made worse off, the head does not act wrongly. This is a counterintuitive result.

Defenders of the harm-based account might offer a number of responses. The first, which we can call the “coincidence response,” appeals to the fact that it is “entirely coincidental that the Jewish university lecturer ended up better off and the head of the university presumably had good reason to think that the result would have been different.”Footnote ¹² What matters to wrongness is not the consequences that in fact result but the consequences that the agent believed would result or had good reason to believe would result. This response commits the defender of the harm-based account to a belief- or evidence-relative conception of wrongness, but it protects the conclusion that in cases where a person acts maliciously but unforeseeably benefits the victim, she still acts wrongly.Footnote ¹³

The second response, which we can call the “systematic response,” distinguishes between individual harms and systematic harms to which individual harms contribute. Even if the harm involved in any individual act of discrimination is close to zero, it may contribute to a systematic harm. As Richard Arneson points out, it may be that “One acts wrongly because one fails to act against this massive wrongful discrimination, and in so doing one contributes to the maintenance of a vicious. . . hierarchy.”Footnote ¹⁴ It may do this in two ways. First, each act of discrimination adds to a large-scale social practice, and so discriminators collectively inflict great harm on disadvantaged groups. Second, each act may increase the likelihood that others will contribute in a similar way. Both of these are plausibly true in Nazi University. Although the employee is not harmed, the employer's actions contribute to a systematic practice that causes widespread harm to the Jewish population. In addition, by contributing to this culture of anti-Semitism, each act increases the likelihood that further discriminatory acts will occur.

The third response, which can be called the “baseline response,” offers a more nuanced method of measuring harm. Lippert-Rasmussen opts for a counterfactual analysis of the relevant baseline for calculating harm. On what he calls the “straightforward account,” a person is harmed by some event, E, if she is worse off now than she would have been had E not occurred. There are problems with this view. It counterintuitively implies that if a person is subjected to a discriminatory act that unexpectedly makes her no worse off than she would have been had the act not occurred, she is not harmed, and therefore the act is not wrong (as demonstrated in Nazi University). Partly in response to this problem, Lippert-Rasmussen canvasses two other options: the “no-discrimination baseline” and the “ideal baseline.” The former holds that we should compare the actual outcome of the discriminatory act to the one that would have obtained if it had not been performed and no one else had performed discriminatory acts in the future. The latter sets the baseline as that in which everyone else acted morally permissibly from the time of the discriminatory act and henceforth.Footnote ¹⁵

None of these responses salvages the harm-based account. To see this, consider the following modification of Nazi University:

Cambridge University evades the coincidence response, the systematic response, and the baseline response. First, the harmlessness of the discriminatory act is not coincidental. This is explicit in the stipulation that Helen discriminates only on the condition that she can use her connections to ensure that applicants are not harmed. If Helen had any grounds to believe that she could not secure a rejected applicant a place at Oxford, she would not discriminate.

Second, the systematic response seems at first sight to have force against Cambridge University. Helen's actions may contribute to a racist culture and a harmful systematic practice, even if her individual actions are harmless. However, Cambridge University can easily be modified to meet this response. Suppose that Helen effectively disguises her activity, such that she makes no contribution to a culture of racism. She does not increase the risk that she or others will discriminate in the future nor promulgate damaging racist attitudes. We can even imagine that through her deceptive efforts, Helen is generally seen as a promoter of racial equality, and the ethnic homogeneity at Cambridge is attributed to other causes. Moreover, she does not contribute to a systematic practice that is harmful overall. Rather, she practices strictly harmless and nonpublic discrimination. This practice, whether at the systematic or individual level, is harmless.

Third, reference to neither the no-discrimination baseline nor the ideal baseline will yield the result that applicants are harmed. This is because there are no further impermissible acts, whether discriminatory or nondiscriminatory, that affect the applicants and that the act of discrimination preempts. This distinguishes Cambridge University from Nazi University, for in the latter further wrongs would probably be committed against the Jewish employee had he not been forced to move to the United States.

However, there is a fourth response that requires more comprehensive treatment. This response threatens the category of harmless wrongful discrimination directly. We can call this the “local-harm response.” It draws our attention to the fact that even if victims of wrongful discrimination end up no worse off overall, they typically suffer local harms. The local harm may be sufficient to render the discrimination wrongful. According to this response, the victims of discrimination in Cambridge University are wronged because they are harmed in one way.

One type of local harm in Nazi University is the violation of the employee's preference not to be fired. As Lippert-Rasmussen notes, the Jewish employee “had a preference for being promoted and by discriminating against him the head of the university frustrated this preference.”Footnote ¹⁶ This version of the local-harm response is easily dealt with. In Cambridge University, it is stipulated that the applicants are indifferent between Oxford and Cambridge and hence, on the preference view, are not harmed as a result of the wrongful discrimination. (We may also imagine, perhaps less plausibly, that the applicants’ desire for a prestigious higher education is so acute that they are also indifferent about whether they realize this goal as a result of discrimination.) However, this is not a complete response to the local-harm objection, because the preference-based view is only one available view about the currency of harm. It may be that the applicant is harmed by being denied a place at Cambridge regardless of her preferences.

We can flesh out this idea in a number of ways. Perhaps the harm consists in an affront to the applicant's dignity or a denial of her equal moral worth. Though a setback to these goods will often cause distress, humiliation, or upset, this will not always be the case, and these sensations and experiences are not necessary for the judgment that the victim is harmed. If this is true, and on the further assumption that attending Oxford is no worse for her than attending Cambridge, Cambridge University involves a compensated harm. Though the case involves compensation, there is an initial local harm present.

To deal with the local-harm response more comprehensively, we can construct further cases that eliminate all local harms. Consider the following:

All three cases involve harmless discrimination. In each case, the discriminatory activity can be divided into two elements: an agent or group of agents deliberates inappropriately, and this deliberation then issues in a discriminatory action. Furthermore, unlike Cambridge University, neither Capricious Teacher nor Racist Voter involves the denial of an opportunity to the victim. They are therefore not cases of compensated harm but genuine instances of harmless action. One worry about Racist Voter is that even if the voter does not in fact cause harm, he increases the risk of harm by casting his vote against the applicant. This distinct factor might make a moral difference. But this possibility is screened out of Racist Voter by the stipulation that the voter knows that his vote will not affect the result. This shows that the harm-based account answers the conditions question incorrectly: a harmful outcome is not a necessary condition of the wrongness of discrimination.

At this point, a defender of the harm-based account might return to the idea of affronting dignity or violating equal moral worth to salvage the judgment that these cases involve harm. It is arguable that the best theory of harm is objective and nonexperiential. A theory of harm is objective in this sense if it allows that a person can be harmed by some event, E, irrespective of that person's relevant mental states. So E might harm a person despite the fact that E does not frustrate her desires; E may harm her even if she positively desires E. A theory is nonexperiential if it rejects the claim that in order for E to harm a person, E must have some impact on her experiences.

Some harms seem to be objective and nonexperiential. For example, it is intuitive that an event that frustrates the achievement of a valuable project is harmful independently of the agent's desires and experiences. Here is an example that demonstrates the plausibility of these claims about harm:

Philosophers will be divided over whether the destruction of the manuscript harms Diane. For the sake of argument, we assume that destroying the manuscript before Diane's colleague can publish it frustrates a very significant achievement. Given this assumption, it is at least arguable that Diane is harmed by the destruction of the manuscript even though she is unaware of it and would be indifferent to it if she found out.

If this is true, might affronts to dignity or violations of equal moral worth also be objective, nonexperiential harms? In Capricious Teacher, for example, might the teacher's racist disinclination to spend time in the company of certain students harm them even if the students are indifferent and unaware?

This response is problematic for two reasons. First, it salvages the plausibility of the harm-based account at the expense of rendering it indistinctive, in the sense that it treats too wide a range of intuitively distinct wrong-making properties as harmful. Although broadening a view to accommodate criticism is often appropriate, this version of the harm-based account seems so broad that it is difficult to conceive of any serious alternative answer to the conditions question. It would threaten the fundamental distinction between exclusively consequence-focused views and nonexclusively consequence-focused views by raising the possibility that all objectionable mental states and/or objective meanings necessarily produce (harmful) consequences. It is doubtful that any defender of the harm-based account would wish to broaden the relevant conception of harm to this degree—this addresses opposing views simply by colonizing them.

Moreover, even if there are objective, nonexperiential harms, we still need an argument that the actions in Cambridge University 2, Capricious Teacher, and Racist Voter constitute such harms. For it seems that the actions in the three cases we outline above affront the victim's dignity or violate her equal moral worth, even though it is intuitive that no harm is done. There seems to be an important difference between plausible candidates for objective, nonexperiential harms, such as frustrating a valuable project, and these cases. Imagine, for example, that a racist millionaire decides to give some money away. He randomly picks names from the telephone directory, deliberately discarding any name to which he takes a disliking. The millionaire practices discrimination but does not interfere with any achievement or goal rooted in the actions of the rejected candidates. It is difficult to see how an objective, nonexperiential harm can be attributed without this connection.

We can intensify the force of the conditions objection by pointing to cases of beneficial wrongful discrimination.Footnote ¹⁷ These are cases of wrongful discrimination in which the victim is benefited by an act of discrimination that involves no local harm.

Consider the following:

In Cambridge University 3, Helen expands the applicants’ opportunities by ensuring that they are offered places at Oxford, thus benefiting them. Moreover, since she does not reject their applications to Cambridge, this case involves no local harm either. Despite benefiting the applicants, Helen's actions remain wrongful. Although there may be differing explanations for this wrongfulness, the most promising is that Helen's actions are wrong because they are motivated by the desire not to spend time around dark-skinned students.Footnote ¹⁸ Helen's act is most clearly wrong when the benefit that she confers on an applicant is only very slight. It is less clear what we should make of cases in which the benefit conferred is much larger. Sufficiently large benefits may be capable of defeating the wrongness of the discrimination. One example might be an admissions officer at a much lower-ranking university who uses her connections to secure places at Oxford for applicants to her university. This is a possibility that we are happy to leave open.

Before we proceed to our second general objection to the harm-based account, there are two important responses that we should briefly consider. The first is to bite the bullet and claim that in all of the counterexamples we offer, the agents in question act permissibly. This response might be bolstered by the observation that although there is no impermissible action, the agents exhibit bad character. There is something morally defective about the agents, manifested in their intentions, their character, or their deliberative processes, but none of these defects renders their actions impermissible. One could add that even if these examples retain some force, any account of the wrongness of discrimination will have some revisionist implications about controversial cases, and so a few intuitive counterexamples cannot count decisively against any particular view.

This response can be made in two ways. Either intentions are not relevant to permissibility in cases of discrimination, though they may be relevant generally, or they are never relevant to permissibility, so a fortiori they are not relevant in cases of discrimination.Footnote ¹⁹ The first version is dubious, since it would be odd if the manner in which intentions affect permissible action is so circumscribed. The onus is on the respondent to explain the material difference between discrimination and other categories of cases. Even if such differences exist, they seem to pull in the opposite direction, suggesting that intentions are particularly relevant to permissibility in these cases. Even Scanlon, who argues that intentions are not generally relevant to permissibility, states that:

The second version of the response is not ad hoc and rests on a deeper debate about the sources of wrongdoing. More specifically, it relies on the general thesis that intentions never nonderivatively determine the permissibility of actions.Footnote ²¹ The problem with this response is that the general thesis on which it depends leads to a wide range of other, highly counterintuitive results. Here are two. First, consider a duress case: A is threatened that unless she robs a post office, her family will be killed. The fact that she has access to a reason that would justify robbing the post office does not make this act permissible if she commits the robbery simply because she enjoys the thrill of crime. Second, consider an overdetermination case: A and B both wish to poison C. Two doses together will kill C quickly, whereas one dose will give C a painful death. A doses C. B then doses C, and in doing so, benefits him by preventing a painful death. However, if B's intention is to kill C, he does not act permissibly (we can add that conversely, C acts permissibly if he doses C in order to avert a painful death).Footnote ²² In the light of these implications, insisting that the agents act permissibly in our counterexamples is not a single bullet to bite, but rather one of a volley of bullets. We highlight these implications to emphasize that in the present discussion, insisting that any view will have at least some revisionist implications is not as reasonable as it first appears.

The second response is to concede that our counterexamples involve wrongful action but to deny that they involve wrongful discrimination. Hellman discusses the example of denying someone a job or a place at school because her name begins with the letter A. She claims that though wrongful and discriminatory, it is not wrongful discrimination. Instead, the wrong consists in the official acting outside her delegated authority or failing to do what she has promised to do.Footnote ²³ Similarly, in Cambridge University, perhaps what is wrong with Helen's actions is that she misuses her authority or breaches a promise to her employer, not that she wrongfully discriminates.

This response is vulnerable to three objections. First, the two cases are disanalogous on Hellman's own terms because in her example, “the source of the wrongfulness does not have anything to do with failing to treat each person as a person of equal moral worth.”Footnote ²⁴ This is not true of Cambridge University and its variations, because Helen acts on the basis of explicit racial prejudice. In a variation of Hellman's case in which candidates with names beginning with the letter A are rejected because of a belief that they have lesser moral worth, the act would be wrong for the same reasons (although the presence of widespread historic and systemic injustice is no doubt an aggravating factor in the racial case). In the original case, we assume the motivation to reject the candidate is, though capricious, not based on a belief in lesser moral worth.

Second, it implies that wrongful harmless discrimination cannot be practiced where professional responsibilities do not arise. This is hard to believe. We offer above the example of the racist millionaire who discriminates against certain groups when giving his money away. The millionaire practices wrongful discrimination even though he lacks professional obligations to an employer. Given the structural similarity between this case and Cambridge University, it is reasonable to conclude, by parity of reasoning, that the latter also involves wrongful discrimination. Perhaps there is an alternative to professional duties that is more plausible than the motivational story we have told, but if so the onus is on the respondent to identify it.

Third, even if all of this is rejected, these wrongs are not mutually exclusive. Helen may violate her professional responsibilities and practice wrongful discrimination. In fact, she plausibly does this by practicing wrongful discrimination. To see this, note that the response misidentifies the victims of Helen's actions. Though she may wrong her employer by breaching a promise or contractual term, she also wrongs the applicant, and this cannot be explained by reference to Helen's professional responsibilities alone.

THE EXPLANATION OBJECTION

The arguments of the previous section refute the claim that a harmful outcome is a necessary condition of wrongful discrimination. But they also entail that the harm-based account's answer to the explanation question is false: when discriminatory acts are wrong, what makes them wrong is not always their harmfulness. Or in other words, if harm is not a necessary condition of the wrongness of discrimination, it cannot explain why discrimination is wrong in every instance. This is the explanation objection.

In response to the explanation objection, a defender of the harm-based account might take one of two options. The first is to maintain that harm is necessary for wrongful discrimination, but acknowledge that it may not always explain that wrongfulness. In so far as this concedes that factors other than harm can explain why discrimination is wrong, this move is attractive. However, if the implication is that in some cases of harmful discrimination, the harm plays no role in explaining its wrongness, then this move is ill-motivated. It is hard to accept that the presence of harm does not play any role in explaining why the act is wrong. This is not how we think about the close connection between harming and wronging generally: if harm is a necessary condition of an act's being wrongful, we would expect it to play some role in the explanation of its wrongness whenever it is present.Footnote ²⁵

The second option is to relinquish the claim that harm is a necessary condition of wrongness and hold instead that when discrimination is harmful, its harmfulness alone makes it wrongful, but there are some cases of harmless wrongful discrimination in which there is some other explanation of the act's wrongfulness. This move embraces a pluralism about the conditions question and explanation question—it accepts that some instances of wrongful discrimination are harmless and, a fortiori, are not wrong because they are harmful, but maintains that when harmful discrimination is wrong, it is wrong only because it is harmful.Footnote ²⁶

The revised harm-based account is capable of generating the conclusion that harmless discrimination can be wrong, but it still fails plausibly to answer the explanation question and in this respect inherits a flaw from its predecessor. The general objection is that harm alone cannot explain why discrimination is wrongful when it is.

In support of this claim, consider the following variation of Cambridge University. This variation is identical to Cambridge University, except that the applicants do not receive an offer from Oxford, and so Helen's discriminatory acts are harmful because they deny applicants the opportunity for a good education. With respect to the conditions question, the revised harm-based account correctly implies that this is a case of wrongful discrimination. However, with respect to the explanation question, it still seems that the wrongness is not explained solely by the harmful outcome of Helen's actions.

There are two points in support of this judgment. First, when we attend to the nature of Helen's wrongdoing, we see that it is noncontingent. The wrongness of the act is not contingent on its consequences. To see this, we can observe that Helen's act is wrong for the same reasons in the original version of Cambridge University. We do not deny that Helen's act in the second variation is more wrongful, since we accept that the degree of harmfulness is an aggravating factor (the “degree question”). The point is rather that harm alone does not explain what makes the act wrong: other wrong-making features are also present. Second, as an epistemic matter, we can ascertain that Helen's act is wrong without any calculation of its consequences. We conjecture that many people, if given a description of Helen's action and the further stipulation that her action may be either harmful or harmless, would arrive at the conclusion that Helen acts wrongly.Footnote ²⁷ These two points—that Helen's act is noncontingently wrong, and that we have epistemic access to its wrongness without information about its consequences—suggest that even the revised harm-based account systematically fails plausibly to answer the explanation question because even the modified version is in this respect an exclusively consequence-focused view.