Ethical Theory and Moral Practice 9: 167–185, 2006.

DOI: 10.1007/s10677-006-9014-x 
C Springer 2006

KASPER LIPPERT-RASMUSSEN

THE BADNESS OF DISCRIMINATION

Accepted: 8 March 2006

ABSTRACT. The most blatant forms of discrimination are morally outrageous and very
obviously so; but the nature and boundaries of discrimination are more controversial, and
it is not clear whether all forms of discrimination are morally bad; nor is it clear why
objectionable cases of discrimination are bad. In this paper I address these issues. First,
I offer a taxonomy of discrimination. I then argue that discrimination is bad, when it is,
because it harms people. Finally, I criticize a rival, disrespect-based account according to
which discrimination is bad regardless of whether it causes harm.

KEY WORDS: cognitive discrimination, direct versus indirect discrimination, discrim-

ination, harm, hierarchical versus non-hierarchical discrimination, moral status, racism,
(dis)respect, sexism, valuation versus non-valuation-based discrimination

1. INTRODUCTION

The most blatant instances of discrimination are morally outrageous, and

very obviously so. It is less clear what makes an act or practice1 discrim-
inatory; whether all forms of discrimination are morally bad;2 and indeed
why objectionable cases of discrimination are morally bad. This paper
addresses these issues. First, I define discrimination. I then argue that dis-
crimination is bad, when it is, because it harms discriminatees, and that
alternative, disrespect-based explanations of discrimination’s badness are
unsatisfactory.

2. DISCRIMINATION DEFINED

In many contexts, the term “discrimination” automatically carries a nega-

tive evaluation of the relevant act or practice. In other contexts, the term is
neutral, and to say that someone discriminates is simply to say, roughly, that
1 I use this formula as a convenient shorthand covering the multitude of things that can
be discriminatory, e.g. policies, intentions, attitudes, aims, institutions, objectives, beliefs.
2 I use “bad” as a placeholder for negative moral evaluations. The term is sometimes used

in a more narrow sense allowing that while discrimination is unjust, it need not be bad.
168 KASPER LIPPERT-RASMUSSEN

he distinguishes, or differentiates, leaving it an open question whether this

is bad (Singer, 1983, p. 309, p. 323). This is the sense of “discrimination”
employed, when “ageism” is defined without redundancy as “wrongful
or unjustifiable adverse discrimination on the grounds of age” (Lesser,
1998, p. 87). To mark the difference between these two senses, we can
modify a device proposed by Kwame Anthony Appiah in a similar con-
text. Appiah distinguishes between racialism, which is the not necessarily
objectionable view that the human species is divided into different, bio-
logically real races and racism which is the objectionable view that this
supposedly biologically real division involves a hierarchy of value (Ap-
piah, 1990, pp. 4–5). Now, discrimination in the sense that interests us
here is discrimination against a socially salient group or particular indi-
viduals qua members of a socially salient group. Hence, with a bit of
linguistic ingenuity we can express the distinction we need by separating
P-based discrimination, e.g. age-, race- or sex-based discrimination, which
involves treating individuals differently depending on their P-properties,
but is not necessarily morally objectionable; and P-ist discrimination, e.g.
ageist, racist or sexist discrimination, which involves treating individuals
differently on the basis of their P-properties in a morally objectionable
way. This terminology allows us to say, for example, that proponents of
affirmative action for women aim to correct sexist discrimination through
sex-based discrimination. Obviously, in this essay, it is P-based discrimina-
tion that I have in mind when I ask “Under what conditions is discrimination
bad?”
The general character of discrimination (for and against) is this:

X discriminates against (in favour of) Y in dimension W iff: (i) X treats Y differ-
ently from Z (or from how X would treat Z, were X to treat Z in some way) in
dimension W; (ii) the differential treatment is (or is believed by X to be) disadvan-
tageous (advantageous) to Y; and (iii) the differential treatment is suitably explained
by Y’s and Z’s being (or believed by X to be) (members of) different, socially salient
groups.

Let me fill out the details of this formula. First, the variable ‘X’ ranges over
both individuals and groups of individuals, as well as super-individuals
such as governments, private companies, and social structures. Obviously,
‘Y’ and ‘Z’ range over (groups of) persons and human beings as well.
While it is not essential to my argument, I shall assume that ‘Y’ and ‘Z’
may also range over non-human animals – so that, for example, speciecism
may be said to involve discrimination against animals.3

3 There are many forms of discrimination and discrimination-related harm from which

animals cannot suffer, e.g. those involving lowered self-esteem. However, this does not
imply that speciecism is not a form of discrimination.
 THE BADNESS OF DISCRIMINATION 169

Second, the formula links discrimination to a specific dimension, W,

because discrimination is not an ‘all-or-no-dimension’ affair. Often dis-
crimination involves differential norms for members of different groups,
and those who fail to comply may be discriminated against in ways that
those who comply do not (which is not to deny that complying and non-
complying members share the predicament of being discriminated against
in other dimensions).
Third, the formula refers to discrimination for and against for different,
socially salient groups for the following reason. An employer might be
more inclined to hire applicants with green, rather than brown or blue,
eyes. This idiosyncrasy might not amount to discrimination in the sense
that interests us here, even though, obviously, the employer differentiates
between different applicants (Wasserman, 1998, p. 807).4 This is not to deny
that such idiosyncrasies can be as bad as, and reflect as corrupted a character
as, genuinely discriminatory acts. It is just that in the great majority of cases
they will not seriously harm the disadvantaged party, precisely because of
their idiosyncratic nature. If we confine ourselves to the focus on socially
salient groups this also explains why we do not talk about discrimination
against non-family-members, unqualified applicants or the undeserving.
While it is true of each of us that we are not members of most families,
there is no group of people made up of non-family members. Similarly,
everyone is unqualified for some jobs, although, arguably, the category of
people who are unqualified for any job has no members.5 Deservingness
is not a socially salient feature, because, generally speaking, it is very hard
to tell how deserving a person is. A group is socially salient if perceived
membership of it is important to the structure of social interactions across
a wide range of social contexts. Having green eyes is obviously irrelevant
in almost any social context, whereas an individual’s apparent sex, race,
or religion affect social interactions in many social contexts. While it is
somewhat unclear when a group is socially salient, this is not a flaw with
my formula. It simply reflects the fact that the contours of our concept of
discrimination are somewhat fuzzy.6

4 See, however, Matt Cavanagh (2002, p. 156) on “discrimininating against [people] on

the basis of the number of vowels in their name”.

5 In a society divided into a few enlarged families, or clans, where people are treated

differently depending on family-membership, we might well approach a situation in which

differentiating between people on the basis of family-membership involves discrimination
to the extent that clan-membership is a socially salient group identity. (Of course it is likely
that clan-membership would be socially salient in that way.) A similar point may apply to
a group of people who are literally unqualified for (virtually) any job where membership
of the group is easily detected.
6 Some consider the use of genetic information by insurance companies to discriminate

against people with the relevant kinds of bad luck in the genetic lottery. Since these people
170 KASPER LIPPERT-RASMUSSEN

Fourth, the formula requires the relevant differential treatment to be

suitably explained by membership of different socially salient groups be-
cause, undeniably, there are cases where X treats Y differently from Z, this
is (believed by X to be) disadvantageous to Y, Y and Z are members of
different, socially salient groups, and yet this differential treatment does
not constitute discrimination. In all likelihood, it is a structural feature of
OPEC’s oil-pricing policy that it is advantageous to some ethnic groups and
disadvantageous to others, but to the extent that differential membership
of ethnic groups is not suitably explanatory of this policy, we would not
say that the policy involves ethnically-based discrimination (Boxill, 1978,
p. 257; Garcia, 2003, p. 284). Quite what suitable explanation involves is
contentious, witness the debate about structural discrimination. For present
purposes, however, I can put that issue to one side.
Fifth, it is only when someone is treated in what is, or is believed by
the agent to be, a disadvantageous way that we speak of discrimination
against. (The reverse point applies to favourable discrimination.) In prin-
ciple it is possible for discrimination to be neither for nor against anyone,
because it is simply a matter of treating different kinds of person differ-
ently. Sex-based discrimination could take this form. It would then involve
discrimination against those who do comply with sex roles, e.g. in gender-
policing, although not discrimination against men or women as such. My
formula allows that the disadvantage involved in discriminatory treatment
may be actual but unforeseen by the discriminator, as well as non-existent
but believed to exist by the discriminator. The latter possibility must be al-
lowed so as to accommodate failed attempts of discrimination. The former
must be allowed, because, according to many observers, this is the most
common form discrimination takes.
Sixth, there are many different ways of treating “Y” and “Z” differently.
Initially, one might distinguish between direct and indirect discrimination.
Direct discrimination occurs whenever some of the phenomena described
below take place. What is common to them is that they all involve rep-
resentational items – e.g. desires, beliefs, statements, laws – that refer to,
or otherwise distinguish between, those who are discriminated against and
those who are not in the relevant discriminatory respect. Hence, to directly
discriminate against, say, women you need to represent those you are dis-
criminating against as women. Indirect discrimination occurs whenever an
individual, institution, or practice acts in such a way that the interests of
some individuals are systematically favoured and yet this does not involve

do not constitute a socially salient group, this does not qualify as discrimination on my
account. However, nothing is lost, and perhaps some clarity gained, if we discuss cases
of “genetic discrimination” as cases of unjust, albeit non-discriminatory, use of genetic
information.
 THE BADNESS OF DISCRIMINATION 171
direct discrimination. For instance, many observers would probably say that
much of the employment discrimination against women that takes place
nowadays is indirect – that women are discriminated against not primarily
because men prefer to hire men, but because of the way the labour-market
is set up, and the connected difficulty of reconciling the demands of one’s
job with being pregnant and giving birth to children (MacKinnon, 1987, p.
37, p. 44).7
Direct as well as indirect discrimination may involve structural elements,
whether of an informal character, e.g. sex-based family norms or sex-based
cognitive schemes,8 or of a formal character, e.g. laws forbidding women
from driving cars. Direct discrimination divides into cognitive and non-
cognitive discrimination.9 Direct cognitive discrimination occurs when a
subject’s beliefs are formed in a way that is biased towards certain people,
e.g. when comparable evidence for men’s and women’s ability to become
good managers will lead the subject to believe that men are capable of
becoming good managers but will not persuade the subject that women
are similarly capable, or when this bias manifests itself in behaviour.10 For
obvious reasons, subjects are rarely aware that they are cognitive discrim-
inators, although this is possible.11
Similarly, non-cognitive discrimination obtains when a subject’s desires
and values are biased, or when this kind of bias manifests itself in be-
haviour.12 The bias is transparent when the content of the subject’s desires
and values refers to the group that he discriminates against, e.g. when the
subject prefers the company of members of his own race to members of

7 Unlike me, some prefer to say that such cases involve oppression, exploitation, or
simply injustice, but not discrimination, e.g. (Young, 1990, p. 195). These differences need
not reflect anything other than a terminological variation.
8 See note 10 for an indication of what I mean by “sex-based cognitive schemes”.
9 This distinction applies directly only to cases of discrimination in which the relevant

representation is a mental state. However, even non-mental representations involve mental

states in an indirect way, and for that reason one can also, in those cases, distinguish
between cognitive and non-cognitive discrimination. Thus a discriminatory law that does
not recognize the possibility of rape in marriage involves mental states in the sense that
it expresses or evinces the mental states of those who proposed and successfully voted in
favour of the law.
10 I say “direct” to allow for indirect cognitive discrimination. Such discrimination may

occur even in the absence of direct cognitive discrimination, e.g. when individuals apply
norms of enquiry in an unbiased way but where the norms themselves are biased. For an ex-
ample of structural, cognitive discrimination based on the differential cognitive framework
of white and black members of the jury, see (Wasserstrom, 1977, p. 598).
11 Reverse, cognitive discrimination can occur when the agent’s desire to see two groups

as identical in certain factual respects skews his evaluation of the evidence in this direction.
Such a case need not involve non-cognitive discrimination.
12 By “desires” and “values” in the present context I have intrinsic as well as extrinsic

desires and values in mind.

172 KASPER LIPPERT-RASMUSSEN

other races. Non-cognitive bias is non-transparent when this is not the case,
e.g. when the subject finds certain ways of avoiding taxation worse than
others and the explanation of this is that this activity is mostly engaged
in by members of a certain immigrant community while comparable ac-
tivities that he judges less bad are mostly engaged in by members of the
non-immigrant majority to which he belongs.
Direct discrimination need not involve negative attitudes toward the
discriminatees, as is shown by cases involving reaction-qualifications. Thus
a shopkeeper may discriminate against black applicants, not because he
dislikes them, but simply because he wants to earn a profit and, justifiably
or not, thinks that, given the racist bias of many of his customers, he is
more likely to do so by hiring white assistants. Here, of course, his desire
not to hire blacks reflects no intrinsically racist attitude.
A distinction cutting across the last one is that between valuation-based
and non-valuation-based discrimination. Discrimination is valuation-based
if, and only if, it satisfies either of the following two conditions: either it
reflects the subject’s view that X’s interests count for more, morally speak-
ing, than Y’s; or it reflects the subject’s view that, while their interests count
equally, X and Y should not relate to each other in the way they should relate
to members of their own group, e.g. as in the case of those who subscribe
to traditional family values and, accordingly, hold that while the interests
of men and women count equally, there is a gender-determined division
of labour. I shall refer to the first variety of valuation-based discrimina-
tion as hierarchical and the second as non-hierarchical. Discrimination is
non-valuation-based when the subject (i) judges that the interests of the
group against which he discriminates are no less important than the inter-
ests of the group against which he does not discriminate (almost always
the group to which he himself belongs), (ii) does not judge that members
of the relevant groups should not relate to one another in the way that they
relate to members of their own group, and yet (iii) prefers to marry, em-
ploy, accompany etc. people with whom he shares (or does not share) his
own group identity. In this case the discriminator has brute discriminating
desires. A brute discriminating desire is hierarchical when it is a desire that
some group will become, or remain, better off than others.
Valuation-based and non-valuation-based discrimination are often con-
fused. Yet, it is important to see that they differ. The way in which they
do so depends on the account of valuation we accept. On an influential
account, valuation can be explained in terms of second-order desires, i.e.
desires with the desirer’s own desires as their objects (Lewis, 1989; Frank-
furt, 1988). Appealing to this account, we have a case of a valuational
non-racial discriminator but non-valuational racial discriminator in some-
one who desires to consort with persons of his own race and yet has a
second-order desire not to act on this desire.
 THE BADNESS OF DISCRIMINATION 173
So much for explaining the formula. Before proceeding to discuss why
discrimination in the above sense is wrong, when it is, I want briefly to
explain three implications of my formula. The first is that discrimination
essentially involves (potentially, at least) treating some individuals dif-
ferently from others. This might be denied on the following ground: to
discriminate against a person it is sufficient to allow one’s treatment of that
person to be guided by facts about him that are irrelevant, or to fail to allow
one’s treatment of him to be guided by facts about him that are relevant.
Thus, a person who complains that he has been a victim of discrimination
when his application for a certain job is turned down on grounds of his race
need not retract his complaint when he learns that members of all races
are equally likely to have their applications turned down on racial grounds
because racists of different colours sit occasionally, and with the same fre-
quency, on the hiring committees. In response, I prefer to say that while
he can complain that, on this particular occasion, he was discriminated
against and that the decision-procedure is a bad one in that it allows irrele-
vant factors to determine who gets hired, he cannot complain that the pro-
cedure was discriminatory, since a procedure cannot discriminate against
everyone.13 In this sense the example is misleading because its target is
unclear.14
Second, reflexive discrimination is possible. It is possible for discrim-
inators to be members of the group members of which they discriminate
against. This happens when female employers reject female applicants be-
cause of their sex. It is even possible for a person to discriminate against
himself, as when an African-American living under Jim Crow laws would
train himself to be deferential in the presence of whites to survive. This is
worth pointing out, I believe, because some of the most damaging forms of
discrimination are those in which discrimination is, so to speak, internal-
ized by the victims of discrimination and involve some sort of abhorrence
of what one is.
Third, discrimination need not be asymmetric. It is possible for someone
to discriminate against others who in turn discriminate against the first
person, and the dimensions of discrimination in each direction may be the
same or distinct. The view that discrimination is necessarily asymmetrical,

13 A procedure may be discriminatory in the derivative sense that it allows discriminatory

decisions to influence outcomes occasionally where these decisions involve treating some
differently from others. However, to say that a procedure is discriminatory in this sense,
even if the procedure as such is not biased against anyone, is not to deny that discrimination
is essentially comparative.
14 If discrimination were not essentially comparative, we could not say that it is bad

because it incorrectly represents discriminatees as having lower moral status. The denial
that discrimination is essentially comparative, would therefore strengthen my argument
against the disrespect-based account below.
174 KASPER LIPPERT-RASMUSSEN

e.g. because only those who have power over others can discriminate against
them, is problematic for reasons that have been set out by Lawrence Blum
in connection with racism (Blum, 2002, pp. 36–39).

3. THE HARM-BASED ACCOUNT OF THE BADNESS OF DISCRIMINATION

I now turn to the issue of what makes these forms of discrimination pro
tanto bad. By saying that something is pro tanto bad, I mean that, from
a moral point of view, that thing is in one respect regrettable. I do not
mean that it is regrettable or impermissible, all things considered, al-
though of course it may happen to be that as well. Another issue that
I intend to steer clear of here is that between consequentialist and non-
consequentialist accounts of what makes discrimination impermissible
(when it is). Advocates of both accounts can accept the harm-based account
of the badness of discrimination that I shall discuss in this section. Their
disagreement will concern what it is permissible to do to eliminate discrim-
ination, e.g. whether it is permissible to discriminate against some people
to eliminate discrimination against more people when these are the only
options.
In my view:

An instance of discrimination is pro tanto bad, when it is, because it makes the discrim-
inatees worse off.

This rubric covers all the forms of discrimination mentioned in the pre-
vious section. It implies that discrimination is contingently bad. Because
of this, some might object that the harm-based account is about as helpful
as an account of the badness of pushing buttons that says that an instance
of button pushing is pro tanto bad, when it is, because it harms someone.
In response to this challenge, note, first, that, in contrast with the case of
button-pushing, many would say that discrimination is non-contingently
bad. Note, second, that many would say, of paradigm cases of discrim-
ination, that the discrimination, even if it is not necessarily bad, is bad
for reasons independent of any harm caused. Finally, again unlike in the
case of button-pushing, certain common, salient, non-contrived types of
discrimination are bad. Hence, given the dialectical setting of the debate
about the badness of discrimination, the comparison with button-pushing
is misleading.
The harm-based account can be specified in various ways. Thus the
relevant baseline for determining whether a discriminatee has been made
worse off and the dimensions that determine how well off someone is can
both be adjusted. Accordingly, what may at first seem to be an objection to
 THE BADNESS OF DISCRIMINATION 175
the harm-based account will often turn out to be an objection to a particular
version of it.15
Consider first the question of the relevant baseline. One could say that
an instance of discrimination is bad when it makes the discriminatee worse
off than she would have been had she not been subjected to it. This would
imply that when someone who is subjected to discrimination that per-
versely makes her better off than she would have been had she not been
so subjected, the discrimination is not bad. Another possibility would be
to say that an instance of discrimination is bad when it makes the dis-
criminatee worse off than she would have been in a just, or the morally
best, outcome. This account has a similar implication in cases where
the discriminatee is not made worse off than she would be under a just
distribution.
In defending the harm-based account, I want to remain neutral on these
two options. However, before proceeding to the second dimension in which
harm-based accounts may differ, I want to show that the second, moralized
baseline view is much more attractive than it may seem at first. First, almost
always when someone raises a complaint on behalf of a discriminatee, the
complainer alleges that the discriminatee is worse off than he should be.
The moralized baseline motivates this focus. Second, it also explains why
there is a moral asymmetry between reverse and non-reverse discrimina-
tion. Reverse discrimination tends to close the gap between how well off
those who benefit from it are and how well off they would be under a just
distribution. Non-reverse discrimination tends to widen this gap. Third,
harm-based, moralized baseline accounts are compatible with the view
that some of the forms of discrimination that I have identified tend to be
worse, because more harmful, than others. Such accounts may distinguish
between being a discriminatee and being a victim of discrimination, i.e.
a discriminatee that is harmed as a result of being discriminated against.
Not all discriminatees are victims of discrimination and some victims of
discrimination are harmed more than others; and normally, just how much
harm it does to a group to be discriminated against depends crucially on its
level of income, education, and cultural self-confidence. Fourth, suppose,

15 Suppose that in order to explain the badness of a certain type of discrimination in

accordance with the harm-based account one has to appeal to a particular version of that
account; that in order to explain the badness of another type of discrimination one has to
appeal to another version; and that these two accounts are incompatible. If this were so, then
my argumentative strategy would fail. However, as far as the second of the two issues that I
discuss below is concerned, there is no inconsistency involved in multi-dimensional, harm-
based accounts that accommodate all the dimensions in which discrimination may harmful
and, thus, bad. As far as the baseline issue is concerned, it is unlikely that anyone will think
that the explanation of what they see as the badness of different types of discrimination will
require appeals to different baselines.
176 KASPER LIPPERT-RASMUSSEN

contrary to what I argued in Section Two, that the group of unqualified

applicants is a socially salient group. In that case, the moralized baseline
does not imply that discrimination against unqualified applicants (based on
their lack of qualifications) is bad. For suppose, plausibly, that for reasons
of efficiency, justice requires a fit between a person’s qualifications and the
sort of job that he gets. In that case discrimination against an unqualified
applicant may not be bad given a moralized baseline even if it harms him
relative to a situation in which he was not discriminated against. Fifth,
denying that a certain discriminatory act that involves no harm is bad im-
plies neither that the agent cannot be criticized for performing it – since
he might, e.g., have had reason to believe that the act would be bad and,
thus, may be blameworthy for performing it – nor that it is not bad that
the agent has a character that disposed him to perform the discriminatory
act.16 Hence, if a discriminatory act makes someone worse off, but not
worse off relative to how well-off she would be under a just distribution,
we can still criticize the discriminator for his conduct, his moral reasoning
or his character, even if we cannot say that the discriminatory act is bad as
such.17
In adopting a moralized baseline, we need to say how the relevant dis-
tribuendum would be distributed under a just distribution, i.e. should we
aim for equality, maximization of the sum of the distribuendum, maximiza-
tion of the weighted sum of the distribuendum (as the priority view would
imply), and so on. This is of paramount importance in determining the
badness of discrimination on the harm-based, moralized baseline account.
If, for instance, we do not think equality is what we should aim for, then
merely pointing out, say, that on average women are worse off than men
as a result of discrimination will not establish that this is bad. The distri-
bution that maximizes the weighted sum of the distribuendum may be one
in which women (or men) are worse off (Richards, 1982, pp. 124–27). Of
course, there could be good reason to think that this is not the case and
that inequality between men and women is evidence that the weighted sum
of the distribuendum has not been maximized. However, this is different
from saying that inequality between men and women is a criterion for the
existence of bad discrimination.

16 See Hare’s distinction between judgements of rightness of the act, judgements of moral

rationality of the act, and judgements of character of the agent in (Hare 1989, pp. 212–230).
Note that the harm-based account can be widened by substituting “. . .it makes or is reason-
ably expected to make the discriminatee worse off. . .” for “. . .it makes the discriminatee
worse off. . .” This should appeal to those who, unlike Hare, think that wrongness and, thus,
badness is tied to reasonably expected rather than to actual consequences of acts. See the
helpful discussion in (Hooker 2000, pp. 72–75).
17 For a related point about intentions and moral impermissibility, see (Thomson, 1999,

pp. 515–518; Scanlon, 2000, pp. 301–317).

 THE BADNESS OF DISCRIMINATION 177
I now turn to the second dimension in which harm-based accounts may
vary – that is, in respect of how we determine how well-off a person is. Here
there are several sub-dimensions in which answers may vary. First, in some
harm-based accounts what matters is whether some people are made worse
off in terms of how their lives go as a whole as a result of discrimination.
This may seem a natural view to hold. However, some theorists reject it, or
insist that it needs to be hedged or supplemented. Consider ageism. Many
people think this is bad. However, if we adopt the whole life account, then,
on the (optimistic) assumption, that all of us eventually grow old, this form
of discrimination would be one from which we will all eventually suffer.
Hence, in principle we could be ageists and no one need be worse off than
others in terms of how his life goes as a whole. To accommodate the view
that ageism is bad, one could adopt a life-segment version of the harm-
based account, i.e. the view that discrimination is bad if it renders someone
worse off in one or more segments of his life.
Second, in multi-dimensional harm-based accounts, discrimination
might be bad when it does not harm the discriminatee overall, provided that
it harms him in some particular dimension, e.g. social status, legal recog-
nition, income, education, or freedom from subordination and oppression;
and this may enable moralized baseline versions of the harm-based account
to accommodate the following sort of case. Suppose that in a just state there
would be no bull fighting, and that in the present unjust state only men are
allowed to be bull fighters. This constitutes discrimination against women,
and we may think this is bad even if women are not made worse off over-
all (because neither men nor women would be allowed to be bull fighters
in a just state) as a result. Adopting a multi-dimensional account, we can
say that legal discrimination against women bull fighters is bad, because it
subjects women to the harm constituted by lack of legal recognition that
they would not suffer in a just state.
Third, some observers think that how well-off people are in the relevant
sense depends on their level of (opportunity to use, or access to) resources.
Others think it depends on their level of welfare. These views have very
different implications. If we think that a just distribution is one in which
everyone’s life contains the same amount of welfare, a demonstration that,
say, men generally command more resources than women will not in itself
show that women are discriminated against in a way that is morally bad.
Things, of course, would be different if we were concerned with resources
rather than welfare.
Fourth, there is the question of whether the units that may be subjected
to harm are individuals or groups (or both). Some have argued that while
black persons who suffered discrimination in the past and are now dead
cannot be compensated, justice requires reverse discrimination because
justice is also, at least, a matter of the equality or compensation of groups
178 KASPER LIPPERT-RASMUSSEN

over time.18 One challenge that group-based accounts of justice face is to

explain how we identify the groups that are relevant from the point of view
of justice (Sher, 1999, pp. 90–93). Whatever the force of this challenge,
there are both individual-centred and group-centred versions of the harm-
based account of the badness of discrimination, so this dispute is internal
to the family of harm-based accounts.

4. THE DISRESPECT-BASED ACCOUNT OF THE BADNESS OF

DISCRIMINATION

The harm-based account ensures that it is logically possible for discrimi-

nation not to be bad. It says that discrimination is not bad when nobody is
harmed by it. Some think this is a mistake. On their view, discrimination
is necessarily bad, independently of the harms and benefits it brings about.
However, I shall now show that the harm-based account of the badness of
discrimination is not alone in implying that discrimination is not bad in
itself.
According to the disrespect-based account:

Discrimination (against other individuals) is bad because it involves disrespect of the

discriminatees and it is morally objectionable to disrespect individuals.19

An act or practice is morally disrespectful of X if, and only if, it in some way presupposes
that X has a lower moral status than he or she in fact has.20

Like the harm-based account, the disrespect-based account is really a

family of accounts. For instance, while advocates of the disrespect-based
account disagree over the properties in virtue of which we merit respect,
typically they agree, e.g., that enslaving, coercing or discriminating against
people involves disrespecting them. The disrespect account is often embed-
ded in a Kantian moral framework according to which we have an agent-
relative duty not to disrespect other persons. However, while this may in fact
be rare, consequentialist theories can also embody a concern for disrespect.
Since, as I have indicated, I want to steer clear of the wider dispute between
18 For a defence of the latter view, see (Taylor, 1973).
19 Presumably, absence of disrespect does not imply respect, e.g. in certain contexts dis-
playing indifference towards another is neither respectful, nor disrespectful. Some theorists
would say that in addition to our duty not to disrespect others, we have a duty to respect
them, i.e. to recognize their moral status: see e.g. (Alexander, 1992, p. 159). For simplicity,
I focus on the duty not to disrespect here.
20 My definition here is partly stipulative. I do not deny that there are senses of “disrespect”

other than the one I am deploying for present purposes, e.g. disrespect constituted by
behaviour that violates conventional norms of etiquette or role-related norms.
 THE BADNESS OF DISCRIMINATION 179
deontologists and consequentialists, I prefer to use the inclusive evaluative
term “morally objectionable”. Similarly, I use “individuals” and “moral
status” as placeholders that may be filled out differently. Most Kantian
theories would substitute “persons” for the former term and “dignity” for
the latter, but I intend my disrespect-based account to be more general – to
include views according to which we should respect non-persons and views
in which the moral status that gets misrepresented in cases of discrimination
is something other than dignity. The equation of moral status and dignity in
the Kantian sense only results in an unhelpfully narrow sense of disrespect.
Offhand, it would seem possible to disrespect someone by expressing the
view that he has no basis for self-esteem, e.g. because his plan of life is
worthless or because he is unable to realize it 21 or by refusing to treat
him “exclusively on the basis of those aspects of his particular character
or circumstances that are actually relevant to the issue at hand” (Frankfurt,
1999, p. 150).22 Note finally that in speaking about acts and practices that
in some way presuppose that X has lower status than he or she has, I aim
to cover cases in which X’s status is not in the discriminator’s mind at all.
Thus if, for purely egoistical reasons, someone ignores the interests of an-
other person in a way that would be morally permissible only if the second
person’s status were lower than it is, the relevant presupposition may be
said to have been made.23
Disrespect-based accounts cannot explain the badness of several of the
forms of discrimination that I have identified. In fact, it seems that the
moral status account can explain at most the badness of valuational, non-
cognitive discrimination. Indirect as well as non-valuational and cognitive
discrimination need not involve representing the discriminatee as having a
lower moral status than he has. It is possible for someone to have all sorts of
cognitive biases against women and yet judge that they ought to be treated
no differently than men.

21 Rawls, of course, would say that as he defines self-respect, this is a denial of self-

respect: see (Rawls, 1971, p. 440). However, as several commentators have argued, Rawls
is using “self-respect” in a non-standard way here (Eyal, 2005, pp. 201–206).
22 Note that one could modify Frankfurt’s remark to make it compatible with my earlier

point about discrimination being essentially comparative, i.e. one could say that disrespect-
ful discrimination is when X treats Y differently in a certain way from Z in dimension W
on the basis of Y’s character or circumstances that are actually irrelevant to the issue at
hand. However, suppose that X treats Y worse than Z on the basis of Y’s character; that
Y’s character is irrelevant to the issue at hand; but that X does this because if he treated Y
according to properties that are in fact relevant to the issue at hand, he would treat Y even
more badly. Here it seems misleading to say that X treats Y disrespectfully.
23 Presumably, if an account of the badness of discrimination says that an act may

constitute disrespect even if it is not suitably related to the claim that the discriminatee has
a lower moral status, it is not a disrespect-based account in my sense. However, typical
accounts of disrespect do involve some such claim.
180 KASPER LIPPERT-RASMUSSEN

It might be argued that, while this is true, the crucial difference be-
tween the disrespect-based and harm-based account is that non-cognitive,
hierarchical, valuational discrimination is disrespectful and hence bad on
the former account, but need not be bad on the latter. To tackle this re-
ply, I now want to argue that there is a deep underlying problem with the
disrespect-based account.
While the harm-based account of the badness of discrimination implies a
moral symmetry between discriminating against someone and discriminat-
ing in favour of someone, no such symmetry exists on the disrespect-based
account. On this account, the problem with discrimination is not that the
person who discriminates is too respectful of some (those in favour of
whom he discriminates) but rather that this person is disrespectful of oth-
ers (those against whom he discriminates). Being excessively respectful
of some people does not entail being disrespectful of others; and one can
be disrespectful of some people without being excessively respectful of
others. Suppose a discriminator believes one group of people to be a moral
elite, and that this belief logically implies there to be at least one group that
has a lower moral standing. Nevertheless, the discriminator may never have
entertained any thoughts about the logical implications of his judgement.
Indeed, whenever he is being presented with examples of non-members
of the relevant group, he may deny that they should be treated any worse
than members of the group that he thinks of as an elite. 24 In such a case,
the discriminator is not disrespectful of anyone, although what he says
logically commits him to such disrespect. Accordingly, if the badness of
discrimination consists in actually being disrespectful, as opposed to say-
ing something in which disrespect is implicit, discrimination in favour of
a group need not be bad.25
This implication of the disrespect-based account is bound to seem
unattractive: presumably, people who wish to account for the badness of
discrimination in terms of respect will want to say that discrimination in
favour of some (e.g. racial) group is bad. Fortunately, however, there is an
obvious way in which the disrespect account can be revised in response

24 He may still discriminate in favour of the first group, according to my formula, in

that only in the case of the first group does he entertain the thought that they have a higher
moral status. In some cases we may conclude that the discriminator does not believe that
the relevant group has a higher moral standing after all, although he says that this is what
he believes. I doubt, however, that this is what we would conclude in all cases.
25 In making this point I assume that what matters is comparative respect. If instead it is

absolute respect that matters (i.e. not how one is respected relative to others, but how one
is respected relative to the absolute level of respect due to one), my argument would be
strengthened. It is possible to be excessively respectful of X relative to the absolute level of
respect due to X, while being respectful of Y to exactly the extent that is due at the same
time.
 THE BADNESS OF DISCRIMINATION 181
to the problem. According to the revised account, discrimination is bad
because it involves a false representation of someone’s moral status, where
the falsehood need not consist in representing someone as having a lower
moral status than he in fact has and where that person need not be the
discriminatee, i.e. the immediate object of discrimination. In a rich arti-
cle, Larry Alexander proposes an account along these lines.26 Alexander
notices that the form of discrimination that most reject is differential treat-
ment of the members of different groups that manifests a judgement of
differential moral worth or status. Such conduct is wrong for the following
reason: “. . . biases premised on the belief that some types of people are
morally worthier than others are intrinsically morally wrong because they
reflect incorrect moral judgments” (Alexander, 1992, p. 161). If discrim-
ination is bad for this reason, it is bad whether or not it harms anyone.27
Moreover, discrimination against and discrimination in favour of are bad
for exactly the same reason, i.e. that both are premised on a false assump-
tion concerning differential moral worth. This, however, introduces a new
set of complications.
First, incorrect moral judgements about a person’s moral worth can be
absolutely or relatively incorrect (or both). An individual, X, may incor-
rectly consider himself morally more worthy than Y and at the same time
consider himself and Y to be morally more worthy than they in fact are.
Presumably, X’s bias reflects incorrect moral judgement, but what makes
the bias disrespectful is X’s incorrect judgement about relativities, not his
incorrect judgement of absolute moral worth.
Second, the direction of the incorrect moral judgement matters. Suppose
X has a positive bias for Y that is based on his incorrect judgement that Y is
morally more worthy than he is. While X may be disrespectful of himself,
and of others who are like him in relevant respects, he is not disrespectful
of Y. Many will consider this bad, but it is not clear that this amounts to
a case of “wrongful discrimination” as Alexander puts it. Whether it does
depends on whether we have a duty to respect ourselves or others whom we
are in relevant respects identical to. If we have either duty, then there is no

26 Alexander thinks that only some forms of discrimination are wrongful, and that of

those that are, only some are intrinsically wrongful. Alan Wertheimer has defended a similar
view: “. . . a preference is less legitimate if it is based on hierarchical judgments about social
groups” (Wertheimer, 1983, p. 107). It is not clear if Wertheimer’s view would be better
expressed if “unjustifiable” were inserted between “on” and “hierarchical”.
27 Some might argue that being disrespected is in itself harmful to the discriminatee.

Accordingly, it is not possible for discrimination not to harm anyone. Whatever the merits
of this view, it does not collapse the disrespect-based account into the harm-based account.
Defenders of the former are committed to the view that even if, per impossibile, discrimi-
nation were not necessarily harmful to the discriminatee, it would still be wrong, because
it reflects incorrect judgements of moral worth.
182 KASPER LIPPERT-RASMUSSEN

reason in principle, on Alexander’s account, why reflexive discrimination

should be less wrongful than non-reflexive discrimination. If, however, our
duty is only to respect others from whom we differ, then it is unclear why
it is wrongful to misrepresent one’s own moral status.
Let us set aside these somewhat fine points and focus on the more basic
question whether the falsity of the discriminator’s belief concerning the
moral status of the discriminatee explains the badness of his discriminatory
conduct (henceforth: the “differential worth thesis”). To assess this matter
rightly, we also need to set aside two confusing factors. First, harmful
behaviour that reflects a judgement of lower moral status is often more
harmful, all other things being equal, than behaviour that does not reflect
such judgement. It may indeed be hurtful when others ignore one’s interests
by oversight, but most people would find it more hurtful if others were to
ignore their interests out of a conviction that those interests are unworthy
of serious attention. However, this is merely a contingent relation and as
such does not support the differential worth thesis. Moreover, the present
point is one that the harm-based account can accommodate.
Second, when we evaluate the differential worth thesis, we tend to have
in mind cases where the discriminator’s judgement of differential worth
reflects his failure to manifest certain epistemic virtues. For example, the
discriminator may apply epistemic double-standards in evaluating evidence
for moral worth or have a tendency to focus on the evidence and exploit
opportunities for critical reflection. No one today, I think, is epistemically
justified in believing that racial or sexual identity alters one’s moral status.28
However, it is certainly possible for an agent to be epistemically justified in
making an incorrect, differential judgement of moral worth even if there are
no true judgements of differential moral worth that would justify common
forms of discrimination. Presumably, some Cartesians were epistemically
justified in not ascribing moral status to animals. Presumably also, given the
deep disagreements surrounding the matter, there may be people nowadays
who justifiably hold false beliefs about whether foetuses have the same
moral status as mature, rational human beings.
False beliefs about moral worth are intimately connected with the bad-
ness of the act in question. For if an act is motivated by a false judgement of
differential moral worth, someone is presumably being treated in a way that,
for independent reasons, he should not be treated. However, it is unclear
what role is played by the fact that the agent’s judgement of differential
worth is unjustified and by the fact that the discriminator has false be-
liefs about the moral status of the discriminatee. Consider first the issue
of epistemic justification. It would appear that in some cases epistemic

28 See Jeff McMahan on racism in (McMahan, 1997, p. 126).

 THE BADNESS OF DISCRIMINATION 183

flaws reflect moral flaws.29 One possibility here is to say that while acts
based on unjustified, false judgements of differential moral worth are no
worse than acts based on justified such judgements, the character of the
first sort of discriminator is worse, morally as well as epistemically, than
the character of the latter. The alternative is to say that the epistemic flaw
somehow amplifies the moral badness of the discriminatory act in those
cases where an epistemic flaw constitutes a moral flaw. For present pur-
poses, however, this issue can be set aside. Our question, as to what moral
difference it makes that the discriminator has false beliefs about the moral
status of the discriminatee, pertains to cases in which the discriminator is
either epistemically justified in holding these beliefs or not thus justified
but in a way that reflects no moral flaw on his part. Compare the following
two cases.

Case (1): An agent discriminates against animals in a way that reflects his false beliefs
that there is a huge difference in the moral worth of animals and human beings and that
it is permissible to inflict almost any amount of pain on any number of animals to benefit
human beings (e.g. he performs a large number of very painful experiments on animals
to test the tendency of a certain perfume to elicit a mild allergic response in a very small
group of human beings). The agent cannot be blamed for having this view.

Case (2): Like Case (1), except that the agent does not believe that there is a huge
difference in the moral worth of human beings and animals. There are three variants of
this case. In the first (2a), the agent has correct beliefs about the moral status of animals
and human beings. In the second (2b), he overestimates the relative worth of animals.
In both these variants, the agent pursues his aim of obtaining the test results despite his
belief about the moral status of animals. In the third variant (2c), he has no beliefs about
the relative moral worth of animals and human beings.

Now compare Cases (1) and (2a). It is plausible to suggest that if they are
not morally equivalent then what the agent does in (1) is less bad than what
he does in (2a). For, while what the agent did in (1) was wrong, at least he
believed, or had beliefs to the effect, that what he did was not wrong. In
(2a), by contrast, not only were the agent’s actions wrong, he believed them
to be so. The same point applies with even greater force to a comparison
of (1) and (2b), but with less force to a comparison of (1) and (2c). But if
one’s discriminatory activities are less bad when they are accompanied by
an underestimate of the moral status of the discriminatee than they are when
they are accompanied by a correct estimate of the discriminatee’s status,
it follows that discrimination cannot be bad simply because it reflects an
incorrect judgement of moral status. For if it were bad for this reason, the

29 In other cases, epistemic flaws may reflect moral virtues. E.g. if, despite evidence to

the contrary, you continue to believe that your friend is telling you the truth, you may be a
better friend than you would be if your view of the matter simply tracked the evidence.
184 KASPER LIPPERT-RASMUSSEN

presence of the incorrect judgement would make the act worse. This is a
very significant result, because it applies to disrespect-based accounts as
such. Note that it does not imply that the falsity of such belief is not in any
way explanatory of the badness of the discriminatory act. It is merely that
it is not explanatory in the relevant sense. If someone beats a child, falsely
believing that it is not morally impermissible to do so, the falsity of this
belief explains the moral impermissibility of what he does in the sense that
it entails it. However, this is not a fact in virtue of which his beating of the
child is wrong. Rather, it is morally impermissible for him to beat the child
because in so doing he inflicts pain on the child.
While the present paper by no means offers a complete defence of the
harm-based account of the badness of discrimination, it does at least show
that account to be plausible, to accommodate views that are often seen
as incompatible with it, and not to differ from one of its main rivals, the
disrespect-based account, in implying that discrimination as such is not
pro tanto bad. Moreover, if the argument above is right, disrespect-based
accounts are mistaken in trying to ground the badness of discrimination in
the discriminator’s false views about moral status.

ACKNOWLEDGEMENTS

Earlier versions of this paper were presented at the Second Priority in

Practice Workshop at University College London (September 2003), the
University of Aarhus (May 2003), and the Second Meeting of the Nordic
Network for Political Theory (October 2003). I thank Larry Alexander,
Paula Casal, Nir Eyal, Lena Halldenius, Dan Hausman, Nils Holtug, Brad
Hooker, Magnus Jiborn, Jason Lindsey, Mats Lundström, Michael Otsuka,
Roland Pierik, Morten Raffnsøe-Møller, Hans Ingvar Roth, Robin May
Schott, Kristian H. Toft, Torbjörn Tännsjö, Jonathan Wolff, and two anony-
mous referees for useful comments.

REFERENCES

Alexander, L., What makes Wrongful Discrimination Wrong?, University of Pennsylvania

Law Review 141(1) (1992), pp. 149–219.
Appiah, A.K., Racisms, in D.T. Goldberg (ed.), Anatomy of Racism. Minneapolis: University
of Minnesota Press, 1990, pp. 3–17.
Blum, L., Im Not A Racist, But. . . Ithaca: Cornell University Press, 2002.
Boxill, B.R., The Morality of Preferential Hiring, Philosophy & Public Affairs 7(3) (1978),
pp. 246–268.
Cavanagh, M., Against Equality of Opportunity. Oxford: Clarendon Press, 2002.
Eyal, N., Perhaps the Most Important Primary Good: Self-Respect and Rawls, Principles
of Justice, Politics, Philosophy, & Economics 4(2) (2005), pp. 195–219.
 THE BADNESS OF DISCRIMINATION 185
Frankfurt, H.G., Freedom of the Will and the Concept of a Person, in H.G. Frankfurt, The
Importance of What We Care About. Cambridge: Cambridge University Press, 1988, pp.
11–25.
Frankfurt, H.G., Necessity, Volition, and Love. Cambridge: Cambridge University Press,
1999.
Garcia, J.L., The Heart of Racism, in B. Boxill (ed.), Race and Racism. Oxford: Oxford
University Press, 2003, pp. 257–296.
Hare, R., Essays in Ethical Theory. Oxford: Clarendon Press, 1989.
Hooker, B., Ideal Code, Real World. Oxford: Clarendon Press, 2000.
Lesser, H., Ageism, in R. Chadwick (ed.), Encyclopedia of Applied Ethics, vol. 1. San
Diego: Academic Press, 1998, pp. 87–94.
Lewis, D., Dispositional Theories of Value, Proceedings of Aristotelian Society, 63(suppl.)
(1989), pp. 113–137.
MacKinnon, C.A., Feminism Unmodified. Cambridge, Mass.: Harvard University Press,
1987.
McMahan, J., The Limits of National Partiality, in J. McMahan and R. McKim (eds.), The
Morality of Nationalism. Oxford: Oxford University Press, 1997, pp. 107–138.
Rawls, J., A Theory of Justice. Oxford: Oxford University Press, 1971.
Richards, J.R., The Sceptical Feminist. Harmondsworth: Penguin, 1982.
Singer, P., Is Racial Discrimination Arbitrary?, in Jan Narveson (ed.), Moral Matters. New
York: Oxford University Press, 1983, pp. 308–324.
Scanlon, T.M., Intention and Permissibility, Proceedings of Aristotelian Society, suppl. vol.,
74 (2000), pp. 301–317.
Sher, G., Diversity, Philosophy and Public Affairs 28(2) (1999), pp. 85–104.
Taylor, P.W., Reverse Discrimination and Compensatory Justice, Analysis 33 (1973), pp.
177–182.
Thomson, J.J., Physician-Assisted Suicide: Two Moral Arguments, Ethics 109(2) (1999),
pp. 497–518.
Wasserman, D., Discrimination, Concept of, in R. Chadwick (ed.), Encyclopedia of Applied
Ethics, vol. 1. San Diego: Academic Press, 1998, pp. 805–814.
Wasserstrom, R.A., Racism, Sexism, and Preferential Treatment: An Approach to the Top-
ics, UCLA Law Review 24(3) (1977), pp. 581–622.
Wertheimer, A., Jobs, Qualifications, and Preferences, Ethics 94(1) (1983), pp. 99–112.
Young, I.M., Justice and the Politics of Difference. Princeton, NJ: Princeton University
Press, 1990.

University of Copenhagen
Section of Philosophy
Njalsgade 80
2300 Copenhagen S, Denmark
E-mail: [email protected]