10
The paradox of deontology, revisited
ulrike heuer
It appears to be a feature of our ordinary understanding of morality that
we ought not to act in certain ways at all. We ought not to kill, torture,
deceive, break our promises (say)—exceptional circumstances apart.1
Many moral duties are thought of in this way. Killing another person
would be wrong even if it achieved a great good, and even if it led to
preventing the deaths of several others. This feature of moral thinking is
at the core of deontological ethics. But while it is also part and parcel of
ordinary moral thought, it seems puzzling, because it requires an agent
to act (or to abstain from acting) in certain ways, even if by violating the
duty she could achieve a better outcome in terms of the very restriction
that she is required to heed.
If the explanation of the duty not to kill were that it preserves human
life it would be hard to understand why killing is not morally permitted
when it leads to preserving a greater number of lives. Moreover, why
should we not violate a restriction if doing so leads to fewer violations of
that very restriction? There are two questions here. (1) Why not violate
a restriction whose point it is to preserve lives, if doing so leads to
preserving a greater number of lives? (2) Why not violate a restriction if
doing so minimizes the violation of that very restriction? The latter
question concerns what is sometimes called ‘the paradox of deontology’
or the problem of ‘minimizing violations’. The former may concern the
so-called trolley cases, for instance. After all, a trolley hurtling towards
a number of innocent people is not about to violate the moral restriction
1
Many deontologist are ‘threshold deontologists’, meaning either that consequences of
a certain kind—catastrophic ones—must be avoided even at the price of violating restrictions,
or that the restriction can be defeated by consequences whose disvalue is out of proportion with
the particular value that the restrictions protect.
The paradox of deontology, revisited
237
on killing innocent people, as it is not the kind of thing that can be
subject to such a restriction. In this paper I will be concerned only with
(2): the paradox of deontology.
To put it in Nozick’s terms: ‘How can a concern for the nonviolation of C [i.e. some deontological constraint] lead to refusal to
violate C even when this would prevent other more extensive violations
of C?’2
Samuel Scheffler, puzzled by the same idea, put it in terms of ‘agentcentred restrictions’, because the restriction requires an agent (not) to
act in certain ways, contrasting thereby with a requirement that certain
actions not be done. The restriction is thus an agent-relative rather than
an agent-neutral one.3 Scheffler ended up rejecting the idea of agentcentred restrictions, as even on close investigation the puzzle would not
dissolve. Here is his description of the structure of agent-centred restrictions (= ACR):
Suppose that if agent A1 fails to violate a restriction R by harming some
undeserving person P1, then five other agents A2 . . . A6, will each violate
restriction R by identically harming five other persons, P2 . . . P6, who are just
as undeserving as P1, and whom it would be just as undesirable from an
impersonal standpoint to have harmed.4
If there are agent-centred restrictions it can be wrong for A1 to violate R,
even in this situation. Scheffler maintains that ACRs have ‘an air of
paradox’ to them, as they violate what he calls maximizing rationality.5
ACRs appear to clash with the idea of maximizing rationality, because
they require an agent to choose the option that worse achieves her goal
of not harming people.
Let me call this, following common parlance, the paradox of deontology, even though it is not strictly speaking a paradox. Exploring
2
R. Nozick (1974), p. 30.
I will not discuss the distinction between agent-relative and agent-neutral requirements or
reasons in any detail, but I will return to it briefly in Section III, and explain why the agentrelativity of the relevant reasons would not be sufficient for establishing agent-centred restrictions. Just as a general explanation, an agent-relative restriction on killing requires of any agent
that she must not kill; an agent-neutral one requires of any agent to see to it that there be no
killings. A note on terminology: I will use the expressions ‘agent-centred restrictions’ and
‘deontological restrictions’ interchangeably.
4
S. Scheffler (1982), p. 84.
5
Scheffler (1985), quoted from The Rejection of Consequentialism, second edn., p. 143.
3
238
Ulrike Heuer
whether there is really a problem here, and how it might be solved, is the
main objective of this paper.
I will begin by explaining the puzzlement about agent-centred restrictions in more detail. I will then discuss promising as an example of
a kind of action which is thought to be subject to agent-centred restrictions. It will turn out that they are surprisingly easy to explain. The
difficult case is killing—and that should put us on guard. I am going to
argue that the problem is not with the very idea of deontological
restrictions, but with puzzling aspects of the ethics of killing.
i. the paradox of deontology
Some philosophers find deontological restrictions not at all puzzling. If
you are under the spell of consequentialist thinking, or in the grip of
‘maximizing rationality’, there is a puzzle, they maintain. But you only
have to shed those influences and the puzzle will vanish with them.
According to ‘maximizing rationality’, if one of two available options
achieves an agent’s goal better than the other, a rational agent ought to
choose the one that is better. As Scheffler put it:
The core of this conception of rationality is the idea that if one accepts the
desirability of a certain goal being achieved, and if one has a choice between two
options, one of which is certain to accomplish the goal better than the other,
then it is, ceteris paribus, rational to choose the former over the latter.6
Deontological restrictions violate this requirement of maximizing rationality, because, as far as the goal of not killing or harming the innocent
is concerned, five innocent people killed (or harmed) compared to one
killed (or harmed) is clearly the worse option, and therefore the one that
the agent ought to prevent. This presupposes, however, that it is the
impartial value of realizing some goal which serves as the rationale for
the agent-centred restriction. But the deontologist ought not to accept
this. As Scanlon points out, 7 the ‘maximizing rationality’ claim may be
6
Scheffler (1985), quoted from The Rejection of Consequentialism, second edn., p. 143.
Scanlon (1998), discussing Scheffler’s view (pp. 82ff) writes: ‘the argument . . . seems to
rely on the assumption that the negative intrinsic value attached to morally undesirable
actions is impartial—that is to say, that an action that has this disvalue is something that
everyone . . . has reason to prevent’ (p. 83). And further: ‘[I ]f one has adopted a goal (and does
not have reason to reconsider its adoption), then one has, other things being equal, reason to
7
The paradox of deontology, revisited
239
correct, but the deontologist will reject that the rationale of agent-centred
restrictions is that they serve the goal of not killing the innocent. Thus
there is no puzzle.
But the optimism is premature. Even if ACRs are not supposed to be
explained by the goal of not killing the innocent, there has to be
some other explanation.8 We could start by presenting ACRs in terms
of the rights and duties that they involve—a language which is often
taken to be more congenial to the deontologist’s concern. Take the
following description: ‘A person (P1 in ACR) has a right not to be killed,
even if it would prevent the killing of five others. That P1 has a right not
to be killed implies that A1 has a duty not to kill P1, no matter what the
consequences or unless they are catastrophic.’ The deontologist might
claim that we can ground such a right in a more fundamental moral
concern: the respect we owe to others, because of their value and dignity
as persons.9 But how about the respect we owe the five? Of course they
too are deserving of the same respect, and they too have a right not to be
killed. But if A1 were morally permitted to violate P1’s right in order to
help prevent A2’s violation of the rights of P2 . . . P6, then we would have
to interpret P1’s right not to be killed differently. She does not
have a right not to be harmed or killed, unless failing to do so would
be catastrophic (call this the ACR-interpretation of rights). She only has
a right not be harmed or killed, unless failing to do so would prevent
a greater number of similar rights violations (call this the MV-interpretation10 of rights).11,12 Why is the ACR-interpretation the right one?
prefer actions that better accomplish it. Taken as an observation about one common but
nonetheless distinctive element in practical thinking, this claim is quite correct’ (p. 85). But,
Scanlon concludes, it is a misunderstanding to believe that the deontologist is one who has
adopted the goal of preventing a certain kind of disvalue from occurring.
8
I take this to be Scheffler’s point: ACRs cannot be explained by maximizing rationality—
but what then would explain them? Would their explanation, whatever it turns out to be, be
compatible with maximizing rationality?
9
Frances Kamm (2007), for instance, regards the ‘moral status’ of a person as an
expression of her value: ‘ . . . my account highlights an agent-neutral value: the high degree of
inviolability of persons that lies at the base even of an agent-relative duty’ (p.29). Similar in her
‘Moral Status’, in Kamm (2007), chap. 7.
10
‘MV’ for ‘minimizing violations’.
11
There are, of course, other exceptions as well: a person may have no right not to be killed
if the case is one of self-defence, or just war, for instance.
12
Frances Kamm (2007) argues that this would be a lesser right. She claims that what is
permissible or impermissible to do to a person fixes her moral status. The moral status of
rational persons is very high. If it were morally permissible to violate a person’s right not to be
240
Ulrike Heuer
The MV-interpretation of rights is not committed to consequentialism: it does not require that an agent should always perform the action
that produces the best outcome. MV is silent on outcomes in general. It
only requires minimizing rights violations. The MV-interpretation is an
alternative both to ACR and to consequentialism, and it is not obvious
why it ought to be rejected in favour of the ACR-interpretation. This is
another way of articulating what seems puzzling with ACRs: favouring
the ACR-interpretation over the MV-interpretation is, without further
argument, unjustified.
But let us just grant that (for whatever reason) the MV-interpretation
of rights ought to be rejected, and accept the ACR-interpretation.
(In Section IV I will present an argument why this is indeed correct.)
Even so, there is yet another way of explaining why deontological
restrictions seem puzzling. Even if we have an ACR-involving right
not to be harmed or killed, is not there also a duty to help and protect
people from coming to harm if we can? Some philosophers flatly deny
this;13 others claim that such a duty is at any rate weaker than the duty
not to violate the right not to be harmed.14
I will assume that there is a duty to help others and protect them from
coming to harm—whether or not it is weak is a different matter.15 It
killed if it prevents a greater number of rights violations, her status would have to be lower than
it actually is. This argument attempts to show that the ACR-interpretation of rights is correct
and the MV-interpretation ought to be rejected. But I am doubtful about its success, because
it seems to beg the question: A person’s moral status is determined by the moral propositions
that are true of her (for example, that she has an ACR-involving right not to be killed), and the
moral proposition is true because it correctly reflects the moral status of a person. But we do
not seem to have any independent purchase on the value of persons that would allow us to
explain a person’s status (and in turn why respecting it requires her having ACR-involving
rights). We cannot determine a person’s status independently of knowing what her rights are.
Of course, Kamm may nonetheless be correctly articulating our (pre-theoretical) view of the
status of persons, and their value. But there is no argument here why the moral status of
rational beings requires that they have ACR-involving rights. The claim is just that this is how
it is. As far as I can see, Kamm does not show that the ACR-interpretation is correct, but she
presupposes that it is.
13
Perhaps some libertarians such as Narveson (2003). They may, however, like Nozick
(1974), only wish to maintain that the state ought not to enforce this duty, rather than that
there is none.
14
Philippa Foot (1967) writes: ‘[t]o refrain from inflicting injury ourselves is a stricter duty
than to prevent other people from inflicting injury, which is not to say that the other is not
a very strict duty indeed’ (p. 29).
15
While this claim is not uncontested, it is nonetheless common ground between adherents of otherwise very different theories. Compare Peter Singer’s claim (1972): ‘ . . . if it is in our
The paradox of deontology, revisited
241
seems obvious to me that if you can prevent a person from being killed or
harmed at little or no cost to yourself, you must do so. If all it takes to save
a life is to call an ambulance, for instance, failing to do so is a grave moral
fault. But even if the costs are considerably higher, it still seems to me that
you have a duty to help. Furthermore, those who defend ACRs often
accept that we have such duties. Thus, the question is not whether a
bystander has a duty to help the five who are about to be killed by
someone else—I will assume that she does—but whether the reason to
comply with this duty is defeated if it requires killing an innocent person.
A simple consideration supporting the claim that there is at least
some duty to help if one can is the following: the ACR-involving right
not to be harmed is supposed to be grounded in the value and dignity of
persons. It is because persons are of value that they are protected by such
rights. But how could the value and dignity of persons ground a right
not to be harmed, but fail to ground a duty to help and protect?16 After
all, we have such duties with regard to all other things of value. We have
not only a duty not to destroy a valuable artwork, but also a duty to
protect it when it is threatened to be destroyed. It would seem utterly
bizarre to think that while you should not throw acid at Rembrandt’s
Nightwatch, if you see someone else setting out to do so (and you can
prevent it) it is none of your concern. The same holds, mutatis mutandis,
for animals, other artifacts, and perhaps for natural beauty. It seems that
the duty not to harm and the duty to protect normally go hand in hand.
Why would this be different when it comes to persons?
But if there is an obligation to help persons and protect them from
coming to harm, then A1 is not only under an obligation not to violate P1’s
right not to be killed, but she also has a duty to help P2 . . . P6, and it is by no
means obvious how this conflict ought to be resolved. Thus, the very same
value, the special value of persons, gives rise to potentially conflicting
power to prevent something very bad from happening, without thereby sacrificing anything
else morally significant, we ought, morally, to do it’, and an Aid Principle, put forward by
T. M. Scanlon (1998): ‘ . . . if you are presented with a situation in which you can prevent
something very bad from happening, or alleviate someone’s dire plight, by making only a slight
(or even moderate) sacrifice, then it would be wrong not to do so’ (p. 224).
16
Kaspar Lippert-Rasmussen (1996) raises a similar worry about Kamm’s idea of moral
status, grounded in the value of persons: why would the status of persons be determined only
by inviolability, as Kamm maintains, rather than by other considerations as for instance by
their non-sacrificibility (Lippert-Rasmussen’s term)?
Ulrike Heuer
242
duties: the duty not to harm persons and to respect their rights, and the
duty to help them when they are threatened to be killed. The standard
examples for the paradox of deontology all illustrate just such a conflict.
But the ACR-involving view of rights does not solve this conflict,
because it does not address it. It focuses on the conflict between not
violating someone’s right vs preventing rights violations of the same kind
by others, but it is silent on the conflict between not violating someone’s
right and helping others. It may be taken (by some) to entail that
whenever there is a conflict between the duty not to violate a person’s
rights and the duty to help, the duty not to violate the right wins out. But
the ACR view of rights does not provide any support for this conclusion.
(The duty to help simply does not come within its purview.) And surely it
cannot draw on the value of human beings for support either.
The not-violating-rights vs helping conflict cannot be brushed aside
by blaming it on the spectre of consequentialism. It is not wedded to
consequentialism. The problem it poses is not due to any assumptions
about minimizing rights violations, or maximizing the good. It arises
because the respect for value requires both not destroying and protecting what is of value.
A number of solutions to the paradox of deontology have been
proposed over the years. Deontological principles such as the Doctrine
of Double Effect (DDE) or the Doctrine of Doing and Allowing (DDA)
may explain why not violating a right takes precedence over preventing its
violation, and why not harming a person is more important than not
helping her—in both cases irrespective of the numbers involved. But of
course these explanations are only as plausible as the proposed principles
themselves. More recently, Frances Kamm and T. M. Scanlon have
offered different explanations, which deserve to be explored further.17
But I am not going to discuss any of them here. I want to approach
the Paradox of Deontology by taking the discussion away from the all
things considered level of principles, and focus instead on understanding
the pro tanto reasons that are involved. I hope that doing so will help
with better understanding the nature of the conflicts.
In the following section I want to take the problem to a different area:
the explanation of the duty to keep one’s promises. My reason to shift
17
Kamm (2007); Scanlon (2008).
The paradox of deontology, revisited
243
the focus at this point is that I want to clarify whether the problem is
really with the structure of deontological restrictions—whether there
actually is an ‘air of paradox’ to them—or whether the problem is of
a different kind altogether.
ii. deontological restrictions and promise-keeping
The paradox of deontology supposedly arises with any agent-centred
restriction. It is not only the restrictions on, or reasons against, killing
which are thought of in this way, but also the reasons against promisebreaking, deceiving others, or torturing them (and possibly quite a few
others).
Promising gives rise to agent-centred restrictions too. That is, assume
that if agent A1 keeps her promise to P1, then five other agents A2 . . . A6,
will each break their promises to P2 . . . P6. It would be wrong for A1 to
break her promise even in this situation.18 Can we explain why this is so?
I will start with the reasons for keeping one’s promises generally, and
then explain A1’s reasons in a situation in which her promise-breaking
would lead to a better outcome in terms of promises being kept.19
II. 1. Some general remarks on promises and reasons
The most striking feature of promising is that by giving a promise one
makes it the case that one is under an obligation to keep it (provided the
promise is valid). It does not follow, however, that we can conjure up
obligations by mumbling ‘I promise’. That there is an act of promising
is only one condition of a valid promise.20 There also has to be
a promisee who accepts the promise, and the promise must not be
forced or immoral in content. All these are only necessary conditions for
giving a valid promise, and I do not aspire to providing sufficient
ones here.
18
As will become clear below, this is not true as it stands—it will need much qualification.
But let it be our starting point.
19
I am aware that the view of promising that I will set out in this paper is controversial, and
that I cannot adequately defend it here. I hope that it will not be too controversial to warrant
discussion of the conclusions that I shall draw from it.
20
I do not assume that promising actually requires the speech-act ‘I promise’. Depending
on context, many different acts can constitute the giving of a promise. Similarly—and perhaps
even more obviously—accepting a promise does not require the speech-act ‘I accept’.
244
Ulrike Heuer
The ability to make valid promises is a power of rational agents to
change their normative situation.21 By promising to ç you put yourself
under an obligation to ç, and you give the person to whom you make
the promise the right to release you from that obligation. Thus a valid
promise creates both an obligation to keep the promise and a right of
the promisee to release the promisor.22
Having this power is of value to agents in a number of ways. Two of
them seem particularly salient to me. (1) The power to make valid
promises enables an agent to bind herself to a certain course of action,
and furthermore (2) it enables her to structure her relations with others
in a certain way—to empower another person, for instance, by giving
her some control over the situation that she would not otherwise have.23
The reason why promises are binding, in general, is that it is of value to
agents to have the power to bind themselves and to structure their
relations to others in the way promising allows them to do. The reason
to keep any particular promise is, however, simply that one promised.
Having undertaken the obligation to ç I will (other things being equal)
fail to act as I ought to do, if I do not ç. I will call this the contentindependent reason for keeping a promise. It exists because agents have
the power to undertake voluntary obligations, and they have this power
because having it is of value in the two ways sketched above.
21
The explanation of promises that I propose owes a lot to Joseph Raz’s (1977) account.
Pace Shiffrin (2008), it does not—I think—involve a ‘transfer of rights’. For some
reasons why not, see footnote 23.
23
Seana Shiffrin (2008) very persuasively explains how promising can do the latter. I agree
in part with Shiffrin’s account, but not with her claim that by promising the promisor transfers
the right to decide on the matter over which the promise ranges to the promisee. As stated
above, I think the promisor gives the promisee the right to release her from the obligation she
has undertaken. The right to decide what to do (if there is such a right at all) remains with the
promisor. Two reasons for rejecting the rights-transfer view are as follows. (1) The reason for
keeping one’s promise can, on occasions, be defeated if, say, I promised to meet you for lunch
but another person urgently needs my help. If by promising to meet you I had transferred the
right to you to decide where I am going to be at 12 noon tomorrow, I would not have the right
to decide that instead of meeting you I ought to help the person in need. But that is obviously
false. The right I give you is different from the right to decide what I do. (2) But by promising
I also do something else (and more) than giving you the right to decide: if I promise you to be
there for lunch, I am under an obligation to be there. It is not that once I promise, you
decide—as if everything was still up for grabs—but I have now (other things being equal)
a sufficient reason to show up. So, in a sense the issue is settled. Hence, promising is centrally
the undertaking of an obligation and not a rights-transfer, even though the promisee is given
a right as well—the right to release the promisor.
22
The paradox of deontology, revisited
245
If giving a valid promise is undertaking an obligation, there is no
further question why there is an obligation to keep one’s promises, and
I assume that we have pro tanto reasons to comply with our obligations.
The only question is why it is that we have the power to bind ourselves
in this way—but that can be explained by its value.24 The value of
having the power to bind oneself also explains the exceptions: why, for
example, immoral or coerced promises are not valid and therefore not
binding. Coerced promises are invalid because they are not instances of
an agent binding herself, while immoral promises are invalid because
there is no value in having the power to create an obligation to act
immorally.25
24
Norbert Anwander and Jan Gertken raised an important worry about this approach.
Doesn’t it involve a rather objectionable kind of boot-strapping? Perhaps a person can have an
obligation to act in a certain way because it would be good if she so acted. But can she also have
an obligation because it would be good if she had such an obligation? That, of course, is not
what I claimed—but perhaps it brings home vividly why there may be some discomfort. My
claim is that persons have the power to undertake obligations if and only if it is good that they
have this power. I cannot fully defend this claim here. But let me make an attempt to dispel the
worry. A normative power is a person’s ability to change the normative situation (to make it
the case that either she or another person has a reason that they would not have had otherwise).
Provided certain conditions are satisfied, in exercising her power the person ‘creates’ a reason.
But there are other ways of intentionally changing one’s normative situation—not by exercising normative powers. Take an example. By intentionally hurting myself I can ‘create’ a reason
to be taken to hospital. This reason can perhaps be explained in terms of value: if a person is
hurt, there is a reason to take her to hospital, because it is good that she will be treated. I can
intentionally make it the case that the antecedent is satisfied, thereby creating a reason. The
explanation of normative powers is different: Here the idea is that it is good that a person can,
by promising, create an obligation. Thus the claim is not, that if a person promises to ç, she has
an obligation to ç, because it is good that she çs, nor that she has such an obligation because it
is good that she has that obligation. This is different from the hospital example, because there it
is not true that there is a reason to be taken to hospital because it is good that by hurting myself
I make it the case that I am entitled to a hospital bed—it is good that if I need hospital
treatment I am entitled to it. That I can make it the case that I need such treatment has no
value. Normative powers can be explained in terms of values, but not in the way in which other
reasons may be so explained. While values play different roles when it comes to explaining
reasons and explaining normative powers respectively, in neither case is there—as far as I can
see—an objectionable kind of boot-strapping involved. The question is really whether this
value-based explanation of normative powers is the right one. Norbert Anwander (2008)
doubts this, but I do not have the space to discuss his view here. The understanding of
normative powers on which I rely has been advanced by H. L. A. Hart (1994), pp. 43–4 (et
passim.), and developed by Joseph Raz (1977). For a more recent defence of a similar view see
David Owens (2006, and forthcoming).
25
This may raise the question why silly promises (which I will discuss in more detail below)
are valid when immoral promises are not. Is there any value in having the power to create
obligations to do something which is not worthwhile doing? I believe there is. Remember that
one of the points of promising is being able to structure one’s relationships—to empower
246
Ulrike Heuer
There are, however, often additional reasons—reasons in addition to
the content-independent ones—for keeping one’s promises.26 The
promisee may expect me to act as promised, and I will let her down if
I do not keep my promise; or she may have made plans relying on my
acting as promised, and would be inconvenienced and disappointed if
I did not keep my promise; or my acting as promised may simply be
important (to the promisee or otherwise) not because it affects her
relationship to me, or her plans, or her emotions, but because I promised to do something which is actually important and worthwhile.
But such additional reasons are only contingently related to
promising. They are not essential because all additional reasons could
exist whether or not there is a promise,27 and the promise will be
binding whether or not there are those reasons. Promises bind even
when they are foolish, or when the promisee does not care at all whether
or not the promisor keeps her promise, or when there is nothing to be
said for acting as promised (because the promise is silly, for instance).
The content-independent reason to do as promised is independent
of the additional reasons. Even if keeping my promise is not
worthwhile I have a reason to do as I promised. Whatever it is that
I promised (provided the promise is valid) there is a reason to act as
promised, because by promising I undertook an obligation to do so.
If I promised to get you a silly hat I have a reason for getting it, whether
or not you look forward to my offering, or you would cherish it if I got
it for you, or you would be disappointed if I did not. Even if you would
not mind foregoing the promised treat, provided that I promised and
that you accepted my promise, I am under an obligation to get it for
another person, but thereby perhaps also to bond with her. Silly promises can satisfy this role as
much as any other, and therefore they are valid. Tom Hurka suggested to me that immoral
promises are invalid because an agent cannot transfer the right to decide whether she should act
immorally to another person, since she does not have this right to begin with. Silly promises are
not affected by this concern. But this explanation depends on the rights-transfer view of
promises which I reject (see footnote 23).
26
Scanlon (1998) seems to assume that the obligation to keep one’s promises has to be fully
explained in terms of what I call ‘additional reasons’.
27
That is, there may be a reason to do something because someone else relies on my doing
it (assuming that I know this), even if I did not promise to do so. This is so in cases where
I made my intention to do something known to another person, or when I have firm habits of
doing certain things so that others form expectations based on those habits. (For a more
thorough discussion, see David Owens (2006).)
The paradox of deontology, revisited
247
you. You may have accepted my promise only to humour me or not to
offend me. Be that as it may. There is a reason to keep my promise
which is necessarily related to promise-keeping. It comes into existence
simply by promising, and because promising is the act of undertaking
a voluntary obligation.
While the content-independent reason to keep one’s promise—to
comply with the obligation one has undertaken—is a particularly
stringent reason (as are all reasons to comply with an obligation), it is
by no means absolute.28 Sometimes, breaking one’s promise will be the
right thing to do if the keeping of the promise conflicts with some more
important duty or with a more important reason.
II. 2. The deontological restriction on promising
Let me now take this back to the deontological restriction. Isn’t the case it
describes perhaps a case in point where the reason to keep the promise is
defeated? Ought I to break my promise if I can thereby make Anton (and
his four companions) keep theirs? Like myself, by promising Anton undertook a voluntary obligation to act as promised. If he breaks his promise he
fails to comply with an obligation that he imposed upon himself. Is there
then a reason for me to see to it that he keeps his promise because he would
be failing to comply with his obligation if he did not? Are there other
reasons for me to bring it about that Anton keeps his promise?
Let me begin with the additional reasons for promise-keeping.
Roughly, they fall into four categories:29
28
The reasons to comply with obligations are generally seen as particularly stringent.
I believe that this is right, provided we use ‘stringent’ in a somewhat technical sense; it should
not be understood as denoting ‘important’. I mean by ‘stringent’ only that there is a particular
structure of reasons that accompanies obligations. The reasons to comply with an obligation or
a duty are special in that certain kinds of reasons that would count for or against acting in the
required way (if there was no obligation) are simply irrelevant. Reasons from ‘inclination’, as
Kant may have put it, are an obvious case in point. It does not matter whether I feel like getting
you the silly hat once I have promised you one. While ‘not feeling like doing something’ may
at times be or indicate a reason for not doing it, it does not when an obligation is involved. It is
not that my inclination is ‘outweighed’ by the reason to keep my promise—it is simply
irrelevant. I take it that this is the sense in which reasons to comply with one’s obligations
are particularly stringent, and different from reasons that do not involve obligations. The
reasons are stringent in this sense, even if the obligation is petty (as in the case of the silly
promise).
29
This is really only a very rough classification. The categories overlap, and, if you prefer,
you could regard all of them as outcome-related—distinguishing the outcomes by their
contents. Nothing much turns on this.
248
Ulrike Heuer
(a) Outcome related:
It is (independently of expectations, and so on) important that the
promisor acts as promised.
(b) Promisee related:
The promisee may expect the promisor to act as promised—so the
promisor would let down the promisee if she did not keep her
promise.
The promisee may have made plans assuming that the promisor will
act as promised, and would be inconvenienced if she did not.
(c) Relationship related:
It may undermine trust between the promisee and the promisor if the
promisor breaks her promise.
(d) Promisor related:
A promisor may lose her ability to give valid promises if she breaks
them frequently. (Her promises would no longer be accepted.)
Let us assume that Anton’s promise to Paul is about something important and worthwhile; that Paul relies on Anton, and Anton would let
Paul down badly if he would not keep his promise. Are these additional
reasons reasons for me—the bystander—to see to it that Anton keeps
his promise? Assume that Anton promised Paul to give him a lift to
London to see his mother who has fallen ill (and there is no easy way to
travel there fast otherwise). How does Anton’s promise affect me? If
Anton is about to break his promise, then as far as Paul’s getting to
London is concerned, I could give him a lift instead. I need not see to it
that Anton keeps his promise, but only that the outcome will be
secured: that Paul gets to London in good time. The reason to secure
the outcome—while possibly being a reason for a mere bystander—is
not ipso facto a reason to violate the agent-centred restriction. In particular, it is not a reason to violate the restriction, because doing so would
prevent a greater number of violations of the restriction. It may, on
some particular occasion, be a reason to break my own promise nonetheless. Perhaps I promised to see a friend for dinner, and if I take Paul
to London I will not be able to keep this promise. But it would not be
a breaking of a promise in order to minimize promise-breaking: as far as
The paradox of deontology, revisited
249
the outcome-related reasons are concerned, it does not matter whether
the outcome is being brought about by Anton keeping his promise, or in
some other way.
There is also Paul’s expectation and the possible disappointment.
Should I break a promise of mine if that is what is needed to make
Anton keep his promise, thereby avoiding these kinds of bad consequence? As opposed to outcome-related reasons, the only way in which
Paul’s disappointment could be avoided is presumably by Anton
keeping his promise. Perhaps I could persuade Anton to keep his
promise. Ought I to do so? The disruption of Paul’s plans could provide
a reason to bystanders, but his feeling of disappointment may be no
reason for a bystander at all (at least not in general, special relationships
apart) simply because his feelings would be entirely appropriate, given
that Anton is about to break his promise. And it seems to me that there
are no reasons to prevent a person from having appropriate emotions,
even if experiencing them is not particularly pleasant.
But what about the relationship-related reasons? Should I try to see to
it that Anton keeps his promise to make sure that there will be no
friction in their relationship? As in the emotions case, frictions in
people’s relationships, when they adequately reflect the reasons that
are relevant, are not per se regrettable. They may simply be appropriate.
Finally, there are the promisor-related reasons—reasons to make
Anton keep his promise, because if he does not he would not be trusted
in the future (by Paul, and perhaps even by others). Imagine that Anton
is a friend of mine, or my child, and that he often does not keep his
promises. He runs the risk that no one will accept his promises any more
if he breaks this one too. Perhaps I worry about him in that regard. In
that case, I may have a reason to do something to make him keep his
promise—but whether ‘something’ would be the breaking of a promise
of mine is again contingent, and perhaps unlikely, since my concern for
Anton’s trustworthiness may lose its credibility if I am readily willing to
undermine my own.
Thus the only (additional) reasons that can easily be reasons for
a bystander are the outcome-related reasons. All others could occasionally be reasons for a bystander, but we need to explain why they are
when they are. They are not ipso facto reasons for everyone. And
furthermore, none of the reasons for a bystander is obviously a reason
to break her promise, in order to make Anton keep his. The outcome-
250
Ulrike Heuer
related reasons are reasons to do something to secure the outcome. The
promisee-related reasons are reasons to do something to prevent the
disruption of Paul’s plans, his disappointment and his being let down,
and so on.
Let us assume next that the promise is a silly promise and there aren’t
any additional reasons for keeping it (i.e., the outcome is not worthwhile, the promisee does not care about the promise being kept, and so
on). In that case, as far as the additional reasons are concerned there
obviously is no reason for a bystander to see to it that the promise will be
kept—simply because there aren’t any.
What about the content-independent reason for promise-keeping
then? After all, this is the only reason that is non-contingently related
to the promise. Do I have a reason to bring it about that Anton complies
with this reason (all additional reasons apart)? There is an explanation
why Anton should comply with his reason: he put himself under an
obligation to act in a certain way. But perhaps all the reasons that can be
shared, and where a bystander’s effort to make someone keep his
promise could make sense, are exhausted by the additional reasons
that may or may not be present.
I think that this is in fact so. To make it more plausible, consider the
following (Kantian) approach. If Anton does not keep his promise, one
way of explaining what is objectionable here is that his will is not as it
ought to be. He is not motivated to act as a moral person would be, or—
as Kantians might express it—he is lacking in good will. Perhaps in
some sense the world would be a better place if it had only people of
good will in it. Would I have a reason to do something that makes it
happen that others act out of good will, if I can do so? And is not that
perhaps the same as a reason to make Anton comply with the contentindependent reason to keep his promise? It would, after all, not depend
on the particular content of the promise, or the outcome-related value
of keeping it. Let me take these questions in turn. First, is there a reason
for making it the case that others act out of good will? And second, if
there were such a reason would it (in the promise-keeping case) be the
same as the reason to make them comply with the content-independent
reason to keep one’s promises?
Relating to the first question: we can tempt others, thereby undermining their resolve and bringing it about that they fail to act out of
The paradox of deontology, revisited
251
30
good will. Doing this will often be wrong. If so, isn’t it then perhaps
also true that we should do what we can to strengthen another person’s
will (not only abstain from undermining her resolve) and make her
comply with her obligations? As Kant sees it, whether an action is done
out of good will makes all the difference to its moral worth, but not,
however, to its permissibility.31 Should we neglect our own obligations
to help others to be morally worthy agents and to act in morally worthy
ways? This would not be a reason to see to it that the other person
complies with her obligation, but more narrowly a reason to see to it
that she does so for the right reasons.
We may have reasons not to undermine the general capacity of others
to comply with their reasons, and to support them in acting for the right
reasons. But ultimately an agent cannot be made to act for specific
reasons. She can be encouraged (or discouraged), influenced, and supported. But it is not within the reach of my agency to bring it about that
you act for the right reasons. Therefore, it cannot be the direct goal of
my action to make this happen. The most likely way to at least influence
the reasons for which others act is by reasoning with them. But we do
not always have reason to do this. There are requirements of respect
which draw the boundaries between mere meddling and good advice.
Thus there is no reason to make others act out of good will, because
another person cannot bring this about directly;32 and furthermore,
there are limits to trying to influence others. Sometimes we have reason
to do so, but not in general. If Anton is a mere stranger I still have reason
to hope that he will turn out a person of good will, but it does not follow
that I have reasons to try to do something about it.33
Our second question was: should I violate my own obligation if this
makes it more likely that you comply with yours—that you comply
30
I am grateful to Michael Ridge for pressing me on this point.
Kant distinguishes between actions that are ‘pflichtwidrig’ (‘contrary to duty’) and
actions that are ‘pflichtmäßig’ (‘in accordance with duty’). If an action is ‘pflichtmäßig’, it
can be done ‘aus Pflicht’ (‘from duty’) or out of self-love (GMS 397). An action has moral
worth only if it is done from duty. Julia Markovits has argued recently that moral worth
depends on the reasons for which an action is done: only if it is done for the right reasons does
it have moral worth. Markovits (2010) steers away (rightly I think) from Kant’s focus on duty.
32
I assume here that we do not have reasons to do what is conceptually impossible.
33
I am grateful to Alex Voorhoeve for suggesting to me the distinction between reasons to
hope and reasons to do.
31
252
Ulrike Heuer
with the content-independent reason for keeping your promise? The
content-independent reason is a reason to act as you promised. It is not
a reason to act, for the right reasons, as you promised. (Imagine that I
promised you I would be in my office at a particular time, but I
completely forgot about it. However, I am in my office anyway. I did
keep my promise; I did comply with the content-independent reason;
but I did not act as promised for the right reasons.) We are now
considering whether I should break a promise of mine if by doing so I
could make it more likely that you will keep your promise for the right
reason (accepting that this is not under my direct control, but I may be
able to influence you). Thus the aim of my action is not to make you
comply with the content-independent reason, but to make you act in a
morally worthy way. Therefore, it is not the content-independent
reason that gives me a reason to break my promise, but a quite different
concern anyway. If we furthermore accept that my reasons to improve
the moral worthiness of others are quite limited (beyond in general
trying not to undermine their capacity of acting for the right reasons),
and have to be restrained by respect for the independence of autonomous agents, the Kantian approach does not provide us with a reason to
make others comply with the content-independent reason to keep their
promises. Perhaps it comes closest to explaining a mere concern for
others complying with their obligations, independently of outcomes.
But since it is not possible to make others act in morally worthy ways,
nor always permissible to attempt to influence them, and since the
concern for moral worth is different from the content-independent
reason for promise-keeping anyway, the Kantian approach does not
establish a reason to make others comply with the content-independent
reason.
But if I do not even have a reason to make others act in morally
worthy ways (as far as that is possible), how could I have a reason to
make them comply with the content-independent reason for promisekeeping when nothing is at stake (when no additional reasons are
present)? The only thing that my interference could bring about in
this case is that a promise will be kept, where there is nothing to be said
for the promise-keeping from the perspective of any of the persons
involved. To claim that I have such a reason you would have to argue
that a world with fewer promises broken in it is, in that regard, a better
The paradox of deontology, revisited
253
34
world. I am not sure that I even understand this claim, and I certainly
see no reason to accept it.
If this is right, then it would not be permissible for me to break my
promise when this would lead to fewer occurrences of promise breakings, since in doing so I violate the obligation to keep my promise for no
reason at all.
In general, while you have a reason to keep your promises, I do not
have a reason to break mine so that you keep yours (additional reasons
apart). But even considering the additional reasons, this is still true for
the most part, because the additional reasons are rarely reasons to see to
it that the other person keeps her promise. They are mostly reasons to
secure the outcome (when doing so is worthwhile).
But of course, there can be cases where the additional reasons would
count in favour of breaking my promise to make you keep yours (where
your keeping your promise is the only way to secure the outcome). But
even in those cases, the reason why the obligation is defeated is not that
more promises are being kept (or that fewer violations occur): that fewer
violations occur is no reason at all to break a promise.
Thus, the obligation to keep ones promises turns out to be subject to
ACRs in a quite natural way. It follows from our understanding of the
value of promising that a bystander cannot realize its value. Its value lies
in the power to bind oneself. Another person can perhaps assist with
realizing whatever particular value the keeping of a particular promise
may have, but having the power to bind oneself does not require
assistance—and it would not be enhanced or protected by seeing to it
that fewer promises will be broken.
II. 3. Agent-relative and personal reasons
Yet the content-independent reason is not simply an agent-relative
reason (even though it is that too). After all, it has often been pointed
out that agent-relative reasons give rise to reasons for others to
34
Holly Smith suggested to me that I ignore the value of promise-keeping. In one sense this
is quite right. I do not believe that there is value in promises being kept (unless there is value in
the outcome of acting as promised, or some other additional reasons obtain). But let me
remind you that I explained the reasons that we have for keeping our promises in terms of the
value of having the power to promise. This is the point where value enters into the explanation,
not at the stage of ‘promises being kept’. But this is also why the value that explains our reasons
for promise-keeping cannot be realized by a bystander.
254
Ulrike Heuer
‘promote’ the agent’s compliance with that reason.35 The agent-relative
reason of a mother to care for her child, for instance, may well give
others reasons to support her and help her in her efforts to do so. That is
why agent-relativity in itself cannot help us to solve the paradox of
deontology.36 The content-independent reason to keep one’s promise
does not provide others with any reasons to ‘promote’ one’s keeping of
the promise. Let me call reasons of this kind personal reasons, and
understand ‘personal’ as indicating that there is a reason which is
a reason only for the agent, and that its presence does not give any
reasons to others.
If the content-independent reason to keep one’s promise is a personal
reason in this sense, then it is not so very puzzling why the duty to keep
one’s promises takes the form of an agent-centred restriction. There
simply is no reason to break one’s promises to make another person
keep hers, and violating one’s own obligation for no reason whatsoever
would be impermissible.
Even the case in which I have reason to bring it about that you keep
your promise by breaking a promise of mine is not a case of violating the
agent-centred restriction. It rather is a case of breaking one’s promise,
because doing so prevents some bad consequences when preventing
them is more important than keeping the promise (and cannot be
achieved in any other way).
Kamm raises an objection37 against attempts to explain deontological
restrictions by appealing to agent-relative reasons which may apply to
personal reasons as well. An agent should not violate a deontological
restriction to prevent a greater number of violations of that restriction—
not even if the violation is necessary to keep herself from violating the
restriction. Those who focus on agent-relativity cannot explain why this
35
Nagel (1970), Parfit (1979), Sen (1982), and Korsgaard (1993) all argue (in different ways)
that an agent-relative reason may well give rise to reasons for others to support the agent in her
compliance with that reason. (Korsgaard rejects the label ‘agent-relativity’—not least because
we can ‘share’ even those reasons which are based on personal pursuits.)
36
As opposed to others, McNaughton and Rawling (2006) seem to think that the agentrelativity of a restriction or a reason is sufficient for establishing deontological restrictions. This
seems doubtful to me, because agent-relative reasons may well give reasons to others: agentrelativity allows that A1 has a reason to promote A2 . . . A6’s compliance with their agent-relative
reasons not to violate the rights of P2 . . . P6—which brings back the paradox of deontology.
37
Kamm (1996), chap. 9. Kamm first discusses the issue in Kamm (1984).
The paradox of deontology, revisited
255
38
is so, and personal reasons fare no better. Since the agent’s reasons for
not violating the restriction once, as well as not violating it many times,
are all of the same kind, it seems that—if numbers count—it would be
better if she violated the restriction only once rather than many times.
Kamm suggests that this is the wrong answer, and the agent-relativity of
the reasons involved cannot explain why.
But is it the wrong answer? First, assume that I have given three
promises, all equally important; that is, as far as additional reasons for
keeping them are concerned they are all on a par, but I can keep two of
them only if I break one. It seems to me that in this case I have most
reason to break the one promise in order to make sure that I keep the
two. At any rate, the question seems to be whether, in a case were all the
options are on a par but an agent cannot comply with all her reasons,
she should comply with the greater number of her reasons. The issue
becomes one of aggregation. Secondly, assume that if I break a promise
now, I can keep myself from breaking two promises that I am going to
give in the future. In this scenario it seems possible that I can comply
with all my reasons and not break any of my promises, if I keep the
promise I have already given but abstain from making the future
promises. If, for some reason, I cannot prevent myself from making
those promises in the future, the case reverts back to the earlier one.
Thus, as far as promise-keeping is concerned, Kamm’s claim is false. But
she had, of course, the case of killing in mind when she put forward her
objection, and I will therefore return to it below.
Thus far there is at least one case where the agent-centred restriction
does not look puzzling. The air of mystery that it is supposed to carry
simply did not make an appearance in our discussion of promises (and
we also did not have to appeal to deontological principles such as DDE
or DDA to explain the restriction.)
But does this show that there is no puzzle after all? Or does it show
that promising is a very special case, utterly different from other moral
duties? I believe that promising is not just a special case, but it is
nonetheless true that the killing cases are much more puzzling and
complicated. The puzzle, as we can see now, is not due to the very
38
Actually, the objection works better (if it works) as an objection to personal reasons for
the reason already explained: agent-relative reasons fail to establish deontological restrictions
anyway, because those reasons can provide reasons for others.
256
Ulrike Heuer
idea of agent-centred restrictions. That can be presented in a perfectly
puzzle-free way. The puzzle pertains to the ethics of killing (or harming)
others—or so I will argue.
iii. deontological restrictions and killing
The reasons to keep one’s promises only contingently combine contentindependent and additional reasons; that is, there may be only the
content-independent reason, and no additional reasons at all. Thus, it
is possible that the obligation to keep one’s promise is personal only (in
the sense defined above): it gives no reasons to anyone other than the
promisor to do anything about the promise-keeping.
The case against killing almost always combines two kinds of reason:
reasons against violating another’s right not to be killed, and reasons
against killing39 because it is harmful. There may be cases where the
killing would not be harmful but still involve a rights violation,40 as well
as cases where the killing would not be harmful and not involve a rights
violation (because the person has consented to being killed and thereby
waived her right).41 Thus, the distinction between violating a person’s
right to life and harming her by killing her is not just an idle distinction.
Furthermore, the harm of being killed is at stake in many cases where
there is no rights violation, as when a trolley is hurtling towards
potential victims. And, as we will see, the reasons for preventing harm
are in some ways different from the reasons for preventing rights
violations. Thus, while in the killing cases the rights violation almost
always combines with harming, the two are distinct, as are the reasons
for preventing them.
Distinguishing the two also helps to avoid a certain mistake. If the
case is one of preventing a rights violation, then it involves other human
agents (the violators of the rights) and their decisions. We may then
conclude that avoiding rights violations is the violator’s responsibility
and no one else’s. To borrow Thomson’s memorable phrase: I do not
have ‘to make the villain’s moral record better than it otherwise would
39
40
41
I should probably speak more generally of the shortening or ending of a person’s life.
Cf. Lippert-Rasmussen (2007).
Perhaps some cases of euthanasia.
The paradox of deontology, revisited
257
42
be’. This may be so. Yet there is a reason for me to interfere nonetheless—not in order to prevent the villain from besmirching her moral
record or because I would be responsible for the killing if I fail to
prevent it, but simply to help the victims. That I ought to help the
victims is independent of the fact that the victims are threatened by a
human agent who is responsible for the killings (that is, it is independent of the killing involving a rights violation). The victims of potential
killers have a claim to being helped just as much as the victims of
runaway trolleys or natural disasters. Some think that the claim to be
helped is not a very strong one—but whatever it is, it is not (I think)
further weakened by the fact that in the one case there is a killer
involved.
I am going to argue that the ethics of killing is so vexing and puzzling
because even if the reasons against the rights violation may take the form
of ACRs, because they are personal reasons, the reasons against harming
and the reasons for protecting others from being harmed do not.43
But why think that reasons against violating someone’s right are
personal reasons? Here is a straightforward, intuitive way of making the
case:
Rescuer’s Decision: Imagine you find yourself in a situation where you
could either save a person from being run over by a trolley or you could
save someone from being killed by another person. (Only in the latter
case does the killing involve a rights violation.) Let us assume that the
death of the person involved is equally bad in each case: they are not
going to suffer; they are of comparable age, health, and so on.
There seems no more reason, in this case, to save the one than
the other. As long as you save one of them you have acted as you
42
Thomson (1985), p. 1414. Bernard Williams (1973) also argues that while an agent is
responsible if she kills someone she is not responsible for someone else’s killings. He is,
however, careful not to draw the conclusion that there is therefore a deontological restriction
on killing.
43
The distinction can also help to show what is wrong with one of the defences of ACRs
and deontology in general. It is sometimes said (even if not, to my knowledge, thoroughly
argued) that while the killer is responsible for the killing, you are not. Therefore, you need not
concern yourself with his killing. The mistake is that even though you are not responsible for
the killing, the victim still has a claim on you to be helped—not because you are responsible
but because you can help, and there is a general duty to aid.
Ulrike Heuer
258
44
ought to. Hence, the rights violation does not give you a further or
stronger reason to rescue the person who is about to be murdered. If so,
then the reason not to violate another’s right not to be killed may be a
personal reason (in the sense explained above). The murderer has reason
not to violate her victim’s right, but you have a reason only to help the
victim because of the harm of killing, and not in order to prevent a
rights violation. That the killing occurs as a result of a rights violation
does not make it worse.
Does Rescuer’s Decision suffice to show that reasons against
rights violations are personal reasons—that you must not violate
anyone’s rights, but that you need not intervene in order to prevent
rights violations by another? This depends on whether the only or the
best explanation for the rescuer’s decision (assuming that it is correct) is
that the reasons not to violate someone’s rights are personal reasons. If we
could establish that the reasons not to violate a person’s rights
are personal reasons, the conflict in the killing case would change its
shape slightly. I began with the assumption that in ACR cases there are
four kinds of reasons involved:
1)
2)
3)
4)
A1 has a reason not to violate P1’s right not to be killed.
A1 has a reason not to harm P1.
A1 has a reason to prevent A2 . . . A6’s rights violations.
A1 has a reason to help P2 . . . P6.
Perhaps the rescuer’s decision shows that we can dispense with (3). But
(1) and (2) would still be in conflict with (4). Can we resolve the
conflict in favour of (1) and (2), thereby perhaps vindicating the ACR?
Would it be because (1) and (2) together are weightier than (4)? Would
(1) on its own suffice to show that (4) has been defeated?
44
Scanlon (1998) makes a similar point: ‘ . . . even those who believe that there is a special
prohibition against intentional killing do not think that one must, for this reason, prevent the
murder rather than the accident. The idea that the murder is a much worse thing to have
happen would seem to imply that one must try to prevent it, even if one has a slightly greater
chance of being able to prevent the accidental death. But this seems wrong’ (p. 83). Compare
also the following passage from Nagel (1986): ‘To use an example of T. M. Scanlon, if you have
to choose between saving someone from being murdered and saving someone else from being
killed in a similar manner accidentally, and you have no special relation to either of them, it
seems that your choice should depend only on which one you are more likely to succeed in
saving’ (p. 178).
The paradox of deontology, revisited
259
I will not try to answer this question here. But is it even the right
question on which to fix? You may have been quite surprised to hear
that there are no reasons of type (3). Especially in the political arena, and
regarding certain kinds of rights, it seems that preventing a rights
violation is a very strong reason—perhaps even the only reason—for
a bystander to intervene. Or is it?
Take freedom of expression. Assume that every person has a right of
freedom of expression against all other people,45 and that its justification lies in the value to the right-holders of having freedom of expression (the condition in which they have unimpeded access to the means
of expression). Call this the ground of the right. It seems that there is
a reason to see to it that speakers have the freedom to exercise their
right—that is, that they enjoy the value which is the ground of the right:
they can make their voices be heard publicly. Whether or not there is
an outcome-related reason for exercising their right to freedom of
expression is here once again a contingent matter. Perhaps their ability
to use the freedom sensibly or wisely is limited or non-existent, so that
exercising their right would not contribute anything of value. That is,
we should distinguish between the value of exercising one’s right to free
speech, and the value of having the freedom to do so. My suggestion is
that only the latter is the ground of the right.
In general, in societies like ours, the right is also a right against the
government, and the responsibility of securing freedom of expression
will lie with law enforcement agents. But imagine that there is no
policeman anywhere in reach, and that someone’s right to free speech
is about to be violated. In that case there may well be a reason for anyone
who can do so to act in a way that will prevent the violation of the right.
Does this show that there are type (3) reasons after all? It does not,
because it may well be consistent with the personal nature of the right.
The reason that bystanders have in such cases would be the benefit for
the person concerned of having freedom of expression. That, rather
than the prevention of the rights violation, may require intervention by
bystanders. It is an outcome-related reason, yet the outcome in question
45
Alternatively, it could be a right against the government only. I assume here that it is
both a right against the government and against other people. We do not have to settle this
issue here. The important point for my discussion is only to show that understanding such
a right (in either of these ways) does not involve type (3) reasons.
260
Ulrike Heuer
is not the value of the speech but the value of having the freedom. The
reason for government protection and intervention is the same: it is
securing the benefit of freedom to the person whose right is being
violated that explains why policemen ought to stop the interference.
The reason is not that they are preventing or minimizing rights violations.46 We can explain why there is a reason (or, in the case of
governments, a requirement) to intervene when someone’s rights are
violated without appealing to type (3) reasons.
It follows from this explanation that a bystander’s reasons to intervene may have to take into account how the outcome—securing the
freedom—can best be realized. That is, if by not intervening in one case
but rather in another, it is more likely that the freedom which is the
ground of the right will be secured (perhaps for a greater number of
people)—that is what the bystander ought to do. Thus her reasons are
different from the agent’s reasons against violating the right: the violator
should abstain from violating the right even if by doing so she can
prevent a greater number of rights violations, because each person has
a right against her to free speech.47 But the person whose right is being
violated does not have a right against the bystander to prevent the
violation: there just is a reason to help for everyone (including the
bystander) to secure the ground of the right (the value that the right
protects). If so, the ability to prevent a greater number of rights violations may count from a bystander’s perspective in a way in which it does
not count from the perspective of the person who is about to violate
a right.48 But it is not the minimizing of the rights violation—even in
the bystander’s case—which furnishes the reason, but the ability to
46
Possibly, the police (as well as all others) sometimes have a reason to prevent certain
rights violation in order to publicly assert the value of the right. But here again the reason
would not be preventing the rights violation as such.
47
This is actually oversimplified. It follows from my analysis that the situation is more
complicated. If considering whether to violate someone’s right in order to prevent a greater
number of rights violations, the way the reasons lie (if I am right) is this. An agent, about to
violate someone’s right has (1) a personal reason not to violate the right and (2) an outcomerelated reason not to violate her right (where the outcome is securing the freedom). The
bystander has (3) an outcome-related reason to protect other people’s rights. There just is no
reason to prevent a rights violation in order to minimize rights violations. In the ACR cases,
however, the agent has type (1) and (2) reasons, conflicting with type (3), and—as in the killing
case—whether she ought to violate one person’s right in order to help a greater number will
depend on how to resolve conflicts of this kind.
48
I am grateful to Peter Railton for pressing me on this point.
The paradox of deontology, revisited
261
secure the outcome (everyone’s freedom to speak up in public). Securing this outcome is presumably not simply a function of keeping the
numbers of rights violations down, but may have to do with the severity
(and perhaps other features) of the rights violations.
How far does this get us towards explaining the deontological restriction on killing? It allows us to say the following. There is no reason for
violating a person’s right not to be killed when doing so would prevent
a greater number of rights violations—simply because there is no reason
to prevent rights violations per se. There is only a reason not to commit
them—and that is, after all, what the agent-centred restriction was all
about. So if the argument above is correct, we have established the
agent-centred restriction not only for promises but also for killing.
And once again, there does not seem to be an air of paradox, nor
a puzzle. It is explained by the personal nature of the reasons against
rights violations.
Let me return to Kamm’s objection against this kind of explanation.
According to Kamm, an agent should not violate someone’s right not to
be killed, even if by doing so she could prevent herself from a greater
number of rights violations of the same kind. The personal nature of
reasons against rights violations cannot explain this, because in this case
all the reasons involved are personal. In the promising case I rejected
the claim on which Kamm’s objection rests: the claim that it would be
wrong of an agent to violate a deontological restriction in order to
prevent herself from a greater number of violations of the same restriction. In the current case I am inclined to do the same. The options of
my killing one in order to prevent myself from killing five, and abstaining from killing one and therefore killing five, look to me very much
like: my killing either one or five, and in that case we may be allowed to
aggregate. Killing one is wrong, but it is the lesser wrong if the only
available alternative is killing five.49 Therefore, Kamm’s objection fails: it
is not true that I should not kill one, even if doing so is necessary to prevent
myself from killing five. Presumably, Kamm would contest the equivalence of the two formulations (‘killing one in order to prevent myself from
49
This presupposes that numbers matter. Thomson (2008) endorses the following principle that she takes from Foot (1967): ‘Killing Five Vs. Killing One Principle : A must not kill five
if he can instead kill one.’ She finds aggregation here no more problematic than in rescue cases
(where it has been most discussed).
262
Ulrike Heuer
killing five’, and ‘killing either one or five’)—claiming, perhaps, that there
is (or can be) a crucial difference between simply killing a person and
killing her in order to save others from being killed.50 I fail to see that
difference—and if there isn’t any such difference, the crucial premise on
which Kamm’s objection rests—that an agent must not kill one, even to
prevent herself from killing several—is false.
However, the reason why killing one would be preferable is not that it
prevents a greater number of rights violations (by the agent). Remember, we said that there is no reason to prevent a rights violation per se.
Thus, minimizing rights violations is not the issue. In Kamm’s cases, the
reasons involved are:
i)
ii)
iii)
iv)
A1 has a reason not to violate P1’s right not to be killed.
A1 has a reason not to harm P1.
A1 has reasons not to violate P2 . . . P6’s rights not to be killed.
A1 has reasons not to harm P2 . . . P6.
My claim that the agent ought to violate the one person’s right rather
than the five persons’ rights does not show that A1 should minimize
rights violations after all. Instead, it shows that when facing the option
of violating one person’s or many persons’ rights, numbers make a
difference: not the number of rights violations, but the number of
people being harmed. But, of course, if A1 decides to violate only P1’s
right, and not P2 . . . P6’s, she commits a wrong. She would be justified
in doing so only if she had an excuse or a justification (if, say, she could
not have prevented being in a situation in which she has to violate
someone’s right). Failing that, violating P1’s right is only the lesser of
two wrongs.
If it is true that there are no type (3) reasons, there is a deontological
restriction on killing in this sense: an agent ought not to violate
a person’s right not to be killed in order to minimize the number of
rights violations of the same kind. Yet, I suspect that this result does not
deliver all that a deontologist would have hoped for. I have not shown
that it would be wrong to kill one person in order to prevent another
person from killing five. I have only shown that the minimizing of rights
50
She offers, as an example, a case where ‘an agent has set a bomb that will kill five people
unless he himself now shoots one other person and places that person’s body over the bomb’
(1996, p. 242).
The paradox of deontology, revisited
263
violation is not a reason to do so. But there are still type (4) reasons—
reasons to help those in danger. And I am not convinced that reasons of
this kind are particularly ‘weak’. Thus it may be that you ought to kill
the one after all, to help the five.
In this regard the killing cases are different from the promising case.
Regarding promises, we can perhaps say that, as a default assumption,
you ought not to break your promise in order to make another person
keep hers. But there is no such default in the killing cases. Consider the
relevant reasons again. Promise-keeping comprises the following three
kinds of reason:
5) A1 has a content-independent reason to keep her own promise P1.
6) A1 may (contingently) have additional reasons to keep her own
promise to P1 (outcome related; due to their relationship; to avoid
becoming untrustworthy).
7) A1 may (contingently) have reasons to secure the outcome of
A2 . . . A6’s promises, or to prevent P2’s disappointment, the loss of
trust between P2 and A2, or of A2’s ability to give valid promises; and
some of those reasons may be due to special relationships between A1
and A2 . . . A6 , or P2 . . . P6 respectively.
The ACR-interpretation of promise-keeping is correct because type (5)
reasons are personal reasons: there is no reason to break one’s promise
just because doing so would make another person keep hers. There
simply are no reasons (conflicting with (5)) of the form:
8) A1 has a reason to see to it that A2 keeps her promise because of the
content-independent reason to keep the promise.
Type (7) reasons are not reasons to prevent the breaking of a promise.
Their explanation has to do with different features that occasionally can
accompany the breaking of a promise. Thus there isn’t always and
necessarily a conflict of reasons in the cases in which the ACR description applies (because there may be no type (7) reasons present at all, or if
there are such reasons, they can be complied with without breaking
one’s promise). If there are type (7) reasons present, and complying with
them requires that one breaks one’s own promise, there is a conflict, and
it has to be resolved by taking into account the relative importance of
the promises and the other considerations. It is not, in those cases,
guaranteed that a bystander could not have most reason to break her
264
Ulrike Heuer
promise in order to make another person keep his, but the explanation
of this case would not appeal to minimizing the violations of the
constraint.
The killing case is different, because there are, in addition to type (1)
reasons, almost always reasons of types (2) and (4) involved. Can the
conflict be resolved—in parallel to the promising case when there is
a conflict—by appealing to the relative importance of the considerations
involved? The answer depends on the answers to two other questions: (i)
How are we to resolve a conflict between violating a person’s right and
helping others generally? (ii) Do numbers count (in conflicts of this
kind)—might the helping be more important, if greater numbers are
involved? I do not have an answer to either of these questions; and at any
rate, dealing with them is for a different occasion.
iv. conclusion
I have shown that there may be a type of reason—which I called personal
reasons—that are similar to certain kinds of agent-relative reasons in
some ways, but different in others. They are reasons not to act in
certain ways which do not provide reasons for others to support the
agent in complying with the reason. If there are such reasons, and if
reasons for not violating deontological restrictions are of this kind, we
have a ready explanation of why they give rise to agent-centred restrictions: there is no reason to minimize the violation of the restriction,
because there simply is no reason at all to prevent others from violating
the restriction. If this is correct, we now also have an explanation why
the ACR-interpretation of rights is correct, and the MV-interpretation
of rights ought to be rejected. If there are no reasons to minimize
rights violations, an interpretation of rights which understood them as
being protected only up to the point at which their violation would
prevent a greater number of rights violations, would be completely
unmotivated.
But this conclusion is more limited than it may at first appear. It
shows only that if we focus on a certain kind of conflict—the conflict
between keeping one’s promise, and preventing another person from
breaking hers (when there are only content-independent reasons
involved), or the conflict between violating a person’s right not to be
killed and preventing someone else from violating such a right—the
The paradox of deontology, revisited
265
answer is clear: there is a reason to keep one’s own promise, but none to
make another person keep hers. There is a reason not to violate
a person’s rights, but none to prevent the rights violations by others.
But that does not give us a conclusive answer what to do, because
sometimes (in the promising case) and almost always (in the killing
case) there are other reasons involved, and they may well count in favour
of violating the restriction.
But perhaps this is as it ought to be. Bernard Williams urged51 that
we want to understand why certain decisions appear so vexing and
difficult. The ethics of killing in particular is troubling and confusing—and a philosophical explanation should not assume otherwise.52
references
Anwander, Norbert (2008). Versprechen und Verpflichten. Paderborn: mentis.
Foot, Philippa (1967). ‘The Problem of Abortion and the Doctrine of Double
Effect’, reprinted in her Virtues and Vices, pp. 19–32. Oxford: Clarendon
Press, 2002.
Hart, H. L. A. (1994). The Concept of Law, second edn. Oxford: Clarendon
Press.
Kamm, Frances (1984). ‘Constraints and You’, paper presented at the APA,
Pacific Division Meeting.
—— (1996). Morality, Mortality, vol. 2. Oxford: Oxford University Press.
—— (2007). Intricate Ethics. Oxford: Oxford University Press.
Kant, Immanuel (1785). Grundlegung zur Metaphysik der Sitten. Quoted from
the Akademie Ausgabe, Vol. IV. Berlin, 1903.
51
Bernard Williams (1973), p. 99.
I presented earlier versions of this paper at the seminar of the Safra Center at Harvard,
the Leeds CEM seminar, the philosophy colloquium at York University, the First Annual
Workshop in Normative Ethics in Arizona, the practical philosophy colloquium at Humboldt
University, Berlin, and the philosophy colloquium at Warwick University. My warmest thanks
are extended to all these audiences for very helpful comments and stimulating discussions. In
particular I would like to thank Keith Allen, Norbert Anwander, Arthur Applbaum, Anne
Barnhill, Anne Burghard, Alix Cohen, Daniel Elstein, Jan Gertken, Stephen Holland, Tom
Hurka, Wouter Kalf, Peter Lamarque, Gerald Lang, Paul Noordhof, Christian Piller, Peter
Railton, Mark Reiff, Tanina Rostain, Thomas Schmidt, Mark Schroeder, Holly Smith, Mark
Timmons, Pekka Väyrynen, and Alex Voorhoeve, for their comments which helped to
improve the paper in many more ways than I can properly acknowledge. I would also like to
thank Christian Miller and Michael Ridge for their comments and suggestions. Finally, I am
especially grateful to Kasper Lippert-Rasmussen and Joseph Raz for extremely helpful written
comments.
52
266
Ulrike Heuer
Korsgaard, Christine (1993). ‘The reasons we can share: An attack on the
distinction between agent-relative and agent-neutral values’, Social Philosophy
and Policy, 10(1): 24–51.
Lippert-Rasmussen, Kasper (1996). ‘The Impermissibility of Minimizing Violations’, Philosophy and Public Affairs, 25(4): 333–51.
—— (2007). ‘Why Killing Some People is More Seriously Wrong than Killing
Others’, Ethics, 117(4): 716–38.
McNaughton, David, and Rawling, Piers (2006). ‘Deontology’, in The Oxford
Handbook of Ethical Theory, ed. D. Copp, pp. 424–58. Oxford: Oxford
University Press.
Markovits, Julia (2010). ‘Acting for the Right Reasons’, The Philosophical
Review, 119(2): 201–42.
Nagel, Thomas (1970). ‘Objective Reasons’, in The Possibility of Altruism,
pp. 90–8. Princeton: Princeton University Press.
—— (1986). ‘Autonomy and Deontology’, in The View from Nowhere,
pp. 164–88. New York: Oxford University Press.
Narveson, Jan (2003). ‘We Don’t Owe Them a Thing!’, The Monist, 86(3):
419–33.
Nozick, Robert (1974). Anarchy, State and Utopia. New York: Basic Books.
Owens, David (2006). ‘A Simple Theory of Promising’, Philosophical Review,
115(1): 51–77.
—— (forthcoming). Shaping the Normative Landscape. Oxford: Oxford University Press.
Parfit, Derek (1979). ‘Is Common-Sense Morality Self-Defeating?’, Journal of
Philosophy, 76(10): 533–45.
Raz, Joseph (1977). ‘Promises and Obligations’, in Law, Morality and Society,
ed. Peter Hacker and Joseph Raz, pp. 210–28. Oxford: Oxford University
Press.
Scanlon, T. M. (1998). What We Owe to Each Other. Cambridge, MA: Harvard
University Press.
—— (2008). Moral Dimensions. Cambridge, MA: Harvard University Press.
Scheffler, Samuel (1982). ‘The Defence of Agent-Centered Restrictions’, in The
Rejection of Consequentialism, second edn, pp. 80–114. Oxford: Clarendon
Press, 1994.
—— (1985). ‘Agent-centered restrictions, rationality and the virtues’, Mind, 94:
409–19; reprinted in his The Rejection of Consequentialism, second edn.
Oxford: Clarendon Press, 1994.
Sen, Amartya (1982). ‘Rights and Agency’, Philosophy and Public Affairs, 11(1):
3–39.
Shiffrin, Seana (2008). ‘Promising, Conventionalism and Intimate Relationships’, Philosophical Review, 117: 481–524.
Singer, Peter (1972). ‘Famine, Affluence and Morality’, Philosophy and Public
Affairs, 1(1): 229–43.
The paradox of deontology, revisited
267
Thomson, Judith Jarvis (1985). ‘The Trolley Problem’, The Yale Law Journal,
94: 1395–415.
—— (2008). ‘Turning the Trolley’, Philosophy and Public Affairs, 36(4):
359–74.
Williams, Bernard (1973). ‘A Critique of Utilitarianism’, in J. J. C. Smart and
B. Williams, Utilitarianism For and Against, pp. 77–150. Cambridge: Cambridge University Press, 1973.