Duress and Duty
ROUGH DRAFT
Victor Tadros1
Around 16th July 1995, Drazen Erdemovic acted as part of a firing squad, killing
around 70 unarmed Bosnian Muslim men from Srebrenica, Bosnia Herzegovina.
These killings were part of the mass execution of Bosnian Muslims at the military run
Pilica Farm in Easter Bosnia where hundreds of men were executed.
The International Criminal Tribunal for the former Yugoslavia (ICTY)
indicted Erdemovic for murder as a crime against humanity. He was indicted after
admitting to a journalist what he had done: ‘because of everything that happened I
feel terribly sorry,’ he said, ‘but I could not do anything. When I could do something,
I did it.’ The reason he could do nothing else, he claimed, was that he was threatened
with being executed along with the victims if he did not participate in their
executions.
The judgement in Prosecutor v Drazen Erdemovic was full of controversy.
Erdemovic was convicted and eventually sentenced to a prison sentence of five years.
The issue of duress was only one of the controversial issues in the case. It is our sole
focus here. The majority held that duress was never a defence to a charge of crimes
against humanity. Two of the five judges, Cassese (presiding) and Li, dissented on the
issue of duress, each holding that duress should sometimes be available as a defence
to crimes against humanity.
Cassese, drawing on the law of a range of jurisdictions, outlined four strict
conditions that he thought had to be met for the defence of duress to be available:2
1) The act charged was done under an immediate threat of severe and
irreparable harm to life and limb.
2) There was no adequate means of averting such evil.
1
Email: [email protected].
2
Para.16.
1
3) The crime was not disproportionate to the evil threatened (this would, for
example, occur in case of killing in order to avert an assault). In other
words, in order not to be disproportionate, the crime committed under
duress must be, on balance, the lesser of two evils.
4) The situation leading to duress must not have been voluntarily brought
about by the person coerced.
The majority, though, followed the common law view that duress is never a defence
to murder.
The moral significance of duress is very important in the context of war. It is
very important in part because in many cases those who participate in the killing of
innocent people during war claim duress. Following WWII, for example, duress was a
common defence to the killing of innocents by members of the Axis powers. Whilst
attempts to use the defence of duress were typically unsuccessful – mainly because
there was often insufficient evidence of a threat having been made – duress was
recognised as providing a defence in principle.
Duress is also important because it has been argued that the fact that soldiers
typically act under duress provides a powerful reason against extending international
criminal liability to all of those who participate in a war on the unjust side. Given that
those who participate on the unjust side cannot point to the cause of the war as a
reason in favour of killing, other reasons must be offered why all of those on the
unjust side ought not to be made criminally liable for their conduct. The fact that
many soldiers are subject to duress is one of the reasons that have been offered.3
My primary aim is to investigate when duress justifies killing in the context of
war. I aim to investigate the moral rather than the legal question. The question is not
whether duress should be available as a defence, in law, to a charge of murder or
crimes against humanity. The question is rather whether and when a person who kills
under duress is morally justified in what she has done. Obviously, the answer to this
question will be very important in evaluating the law. But as always, there may be
good reason for the law to depart from the best moral view. Evaluating the strength of
these reasons is beyond the scope of this paper. Furthermore, my central interest is not
in whether duress can excuse killing. I am interested primarily with duress as a
3
See, for example, J McMahan Killing in War (Oxford: OUP, 2009)
2
justification. For reasons outlined below, this analysis will also be important in
determining whether and when duress can excuse killing.
Here is a roadmap. In Section I, I conceptualize duress as a justification. Section II
outlines two objections to duress as a justification to killing. Section III evaluates one
kind of case where these objections are not decisive – where the killing is a side-effect
of the action that the person is instructed to perform. Section IV evaluates another
kind of case where they are not decisive – where the person killed was doomed to die
whatever the defendant did. It also considers a less familiar variation on this case.
Section V considers more standard cases of duress, suggesting that deontological
restrictions on responding to threats are not as clearly decisive as appears to be the
case. This offers some support to the view implicit in Cassese’s judgement that duress
should sometimes be available even when the victims are intentionally killed and are
not doomed to die.
I.
Duress as a Justification
A person who acts in a way that is pro tanto wrong is not always blameworthy for
what she has done. One reason for this is that the person is a responsible agent - for
example if she is a small child or suffers from a serious cognitive disability. Another
reason is that the person’s conduct is excused - for example if the terrifying threat that
she was faced with undermined her decision-making powers were. A third reason,
which is my focus here, is that the person’s action, although pro tanto wrongful, is
justified all things considered.
Duress might be understood as a justification defence or as an excuse defence.
It is controversial that it should ever be recognised as a justification. Some people
think that duress is best understood only as an excuse.4 But these people don’t deny
that it is sometimes permissible, all things considered, to respond to a credible threat
by acting in a way that is pro tanto wrong. For example, if X threatens D that X will
shoot D’s child if D does not illegally park his car, it is permissible, all things
considered, for D to illegally park. Those who deny that duress can be a defence to
pro tanto wrongdoing do not deny that this is permissible. They rather deny that this
4
See, for example, J McMahan Killing in War 113.
3
is best understood as a case of duress. They use a different defence-label, such as
‘lesser evil’, to refer to this case.5 I will use the label ‘duress’ to refer to it.
Standard cases where D acts under duress, as I will understand it, have the
following features:
1) X threatens D that if D does not do some act v, X will do some act w.
2) D vs in order that X does not w.
3) It is pro tanto wrong for D to v.
4) Were it not for X’s threat it would be wrong for D to v.
If duress is only available as an excuse, in standard cases of duress, not only is it pro
tanto wrong for D to v, it is wrong for D to v all things considered. The reason for this
is that one does not need an excuse for doing what is all things considered
permissible. Hence, duress qua excuse applies only to cases of all things considered
wrongdoing.
As I will understand duress, D may act under duress and yet act permissibly
all things considered in some cases. This will be true in cases where the threat of w is
sufficiently grave to justify X in ving even though ving is pro tanto wrong. These are
the only cases of duress that I am interested in. If duress is to be available as an
excuse, what is needed is an account of how it can be true that the threat of w is
insufficiently grave to justify D in ving, and yet D is not blameworthy for ving. This
might be true, for example, if D’s reasoning is undermined by being placed in a state
of terror as a result of X threatening w.
We should consider duress as a justification prior to evaluating duress as an
excuse. This is because evaluating when duress justifies is important in evaluating
when it excuses. And this is because whether duress is available as an excuse depends
on whether duress is available as a justification. Why should the availability of duress
as an excuse depend on whether duress is available as a justification? The reason is
that the closer D is to being justified for ving, the stronger the reason we will have to
excuse D for ving. We are entitled to an excuse only if our reasoning has been
interfered with in a way that has led us to act wrongly. But we are likely to be excused
5
G P Fletcher and J D Ohlin Defending Humanity: When Force is Justified and Why
(Oxford: OUP, 2008) 123, writing in the context of the Erdemovic judgement.
4
only if our reasoning did not dramatically depart from that which was required. And
that will be so only if our conduct was reasonably close to being justified.
For example, suppose that X threatens D that X will break D’s foot if D does
not kill an innocent person. The defence of duress would not normally be available in
this case. It is not available as a justification – averting the breaking of a foot is
insufficient reason to kill a person. It is also not available as an excuse. The reason for
this is not that having one’s foot broken is insufficiently bad to terrify a reasonable
person. A threat of having one’s foot broken might be terrifying. It is rather that
averting a broken foot is nowhere near important enough to justify killing a person. If
X had threatened D that X would break D’s foot if D did not break the foot of an
innocent person, we would be more inclined to excuse D. That is so even though the
threat that X makes to D is no more terrifying in this case than in the previous case.
The reason is that D’s act is closer to being justified in that case. For this reason, we
are better focusing at first on when threats justify a person in acting in a way that
would be wrong were it not for the threat. This will be important in determining when
duress is available as an excuse.
Evaluating when duress should be available as an excuse is, I believe, a much
harder task overall than evaluating when it should be evaluated as a justification.
Evaluating when it should be available as an excuse requires us to imagine the
psychological impact of duress. Duress in the context of war is often multifaceted and
systematic – many threats may be imposed on a person over a period of time,
reducing her ability to resist any one of them. Her moral psyche may become
corrupted by being subject to constant threats, and by being placed in an environment
where a thin line needs to be walked in order to avoid death, perhaps not only to
oneself but also to those one loves.
In saying this, from the quiet comfort of my desk, I am influenced by Primo
Levi’s words, writing about collaboration in Nazi concentration camps:
Before discussing separately the motives that impelled some prisoners to
collaborate to a varying extent with the Lager authorities, it is necessary
however to declare that before such human cases it is imprudent to hasten to
issue a moral judgement. It must be clear that the greatest responsibility lies
with the system, the very structure of the totalitarian state, the concurrent guilt
on the part of the individual big and small collaborators (never likeable, never
5
transparent!) is always difficult to evaluate. It is a judgement that we would
like to entrust only to those who found themselves in similar circumstances,
and had the possibility to test on themselves what it means to act in a state of
coercion…The condition of the offended does not exclude culpability, and this
is often objectively serious, but I know of no human tribunal to which one
could delegate the judgement.6
Whether collaboration is justified is, I think, easier to evaluate from the armchair than
whether collaboration is excused. It requires less empathy with those who might
collaborate – less understanding of the psychological impact of duress.
Before examining the question of justification further, let me say something
about two features of the standard cases of duress that I am considering. First, duress
is concerned with threats. A threat is sincere if X forms a conditional intention to w if
D does not v. It is credible if X will succeed in executing this intention if D does not
v. In many cases of duress, there will be doubts about both the sincerity and
credibility of the threat. For example, in Erdemovic itself, it is not completely clear
whether those threatening Erdemovic really intended to shoot him if he disobeyed
orders, or if those threatening him would have been able to execute the threat. Perhaps
Erdemovic had some opportunity to escape rather than participating in mass
executions. If so, his claim to a duress defence would be weaker.
Where the threat is less likely to be sincere or credible, we are concerned with
risk. To what extent may D v in virtue of the fact that there is a risk, if he does not do
so, that X will w? Evaluating risks is a tricky business. I leave this question aside
here. I assume that X is completely sincere and credible.
Secondly, we should distinguish between whether some sincere and credible
threat is made and whether the person responded to the threat. The person responded
to the threat if the threat provided the reason for which he acted. There is a question
about whether the duress defence ought to be available in cases where X threatens D
that X will w if D does not v, D vs but he does not do so to avoid X wing. For
example, D might v simply because he wishes to v. Cassese’s criteria for duress,
outlined above, do not include a condition that D is motivated by the threat. But his
judgement does point to the importance of this consideration.
6
The Drowned and the Saved trans. R Rosenthal (London: Abacus, 1988) 28-9.
6
Whether D was motivated by the threat, some might claim, cannot be relevant
to whether D’s act is permissible all things considered. Some might claim this
because they think that intentions are irrelevant to permissibility. If D would be
justified in ving in order to avert the threat of w, these people might claim, D is also
permitted to v for any other reason, including simply his desire to v.
The better view is that duress justifies where it does only if D acts in order to
avert the threat of w. Intentions, in this case, are relevant to permissibility. D cannot
claim that X’s threat to w justified D in ving if D was not motivated by X’s threat to
w. To see the intuitive force of this view, consider the penalty for disobeying orders in
Nazi Germany known as Sippenhaft (kin liability). Sippenhaft warranted not only
killing the person who disobeyed the orders, but also members of the person’s family.
This practice provided the basis of a claim of duress in the Stalag Luft III case
following WWII. In that case, the 18 accused, who were all members of the SS,
participated in the killing of 50 members of the RAF. The court did not believe that
the members of the SS were acting under duress.7 Had they been acting under duress,
though, the court might have granted the defence.
But surely the defence ought not to be made available if those SS members
were not influenced in any way by the threat to them and their families. If they were
fully committed Nazis, the defence of duress ought to have been ruled out, even if
they and their families would have been executed had they not obeyed orders. Were
this not so, we would face the unacceptable spectre that the duress defence would be
very widely available to fully committed Nazis in virtue of the fact that had any Nazi
disobeyed orders other Nazis would have punished them severely.8
At any rate, I will assume, in all of the cases that I consider, that D is
motivated to avert w, and vs only for that reason. We can thus evade the controversial
question whether D’s intentions are relevant to permissibility.
II.
7
Objections to Duress as a Defence to Killing
As Judge Cassese emphasises in Erdemovic para 24, it appears that the court did not
believe that the accused had been subject to duress in that case.
8
For a deeper exploration of this issue, see V Tadros The Ends of Harm: The Moral
Foundations of Criminal Law (Oxford: OUP, 2011) ch.7.
7
As I noted above, subject to terminological disputes, everyone accepts that duress
sometimes justifies acting in a way that is pro tanto wrong. Our question is whether
duress can ever be a defence to killing. English criminal law, which was followed by
the majority in the Erdemovic case, answers this question in the negative. Duress
cannot even provide an excuse to murder in the law of England and Wales.9 It is
unavailable even if D was an accomplice rather than a principal – if D provided P
with a weapon that was used to kill V, and D’s life was threatened, D cannot escape a
conviction of murder. It is also unavailable as a defence to attempted murder.10
This aspect of the law has received a great deal of criticism. It has been
described as ‘too rigid’ on the grounds that ‘there will be occasions where we would
expect even a person of reasonable firmness to be coerced into participating in
murder’.11 Duress ought at least to be made available as an excuse to murder, it has
been claimed, even if killing under duress is never justified.12
In contrast, some may argue that killing under duress is normally so far away
from what is permitted that it ought never to be available as an excuse to murder.
Killing a person to avert the threat of a broken foot ought never to be excused, it
might be argued. Similarly, killing a person to preserve one’s own life in
circumstances of duress ought never to be excused.
To vindicate this view, it would need to be shown that killing a person under
duress is a long way away from being justified, just as killing a person to avert the
threat of a broken foot is a long way away from being justified.
i)
Collaboration
One reason that might be offered for this view is that it is wrong to collaborate with
evil. If D responds to X’s threat, D carries out X’s wrongful intentions. D thus
collaborates with X in fulfilling X’s evil plan. Let us say that D collaborates with X if
9
R v Howe [1987] AC 417, overruling DPP v Lynch [1975] AC 653.
10
R v Gotts [1992] 2 AC 412.
11
A P Simester, J R Spencer, G R Sullivan and G J Virgo Simester and Sullivan’s
Criminal Law: Theory and Doctrine (Oxford: Hart, 2010).
12
A Ashworth Principles of Criminal Law (Oxford: OUP, 2009) 6th edn. 213-4.
8
X forms an intention that D vs in order to advance X’s plan that a wrongful act will be
committed, and D vs, fulfilling this intention. Why might collaboration be wrong?
Here are two arguments that might be offered. First, it might be argued that it
is wrong for D to kill V because D owes it to himself not to collaborate with evil. Let
us call this ‘the agent-relative argument’. Secondly, it might be argued that it is wrong
for D to kill V because D owes it to V not to participate in a plan that treats V as
insufficiently important. Let us call this ‘the victim-centred argument’.
I doubt that the agent-relative argument typically has much force. I think that
the best way to articulate this argument is as follows.13 People build their lives around
distinctive activities, goals and commitments. Collaborating with evil may sometimes
conflict so powerfully with those activities, goals and commitments that the person
will corrupt herself by so acting. She may have a self-regarding duty not to corrupt
herself in this way. Her integrity, to put the point in another way, is threatened by
collaborating with evil, and she may have a duty not to threaten her integrity. 14
This way of developing the agent-relative argument is somewhat unusual in
that it refers to the idea that people may have self-regarding duties not to threaten
their integrity. But, whilst this is not often noticed, this idea seems essential for the
agent-relative argument to work. It is essential for the following reason. Integrity may
provide us with a reason, but not a duty, to refrain from collaborating with evil. If
collaborating with evil threatens our integrity, and our integrity is important to us, we
may be permitted not to collaborate with evil, even where collaboration will benefit
innocents. We can refrain from collaboration if the costs to our integrity of
collaborating are too high.
But even if this argument is successful, it cannot demonstrate that it is wrong
for D to respond to X’s threat. What is needed, then, is not a self-regarding reason not
to collaborate with evil, but rather a self-regarding duty not to collaborate with evil. It
is more difficult to show that any threat that there would be to D’s integrity if he
responds to X’s threat renders it wrong for him to wrong to act on the threat. Why
13
I am grateful, here, to Larry Temkin for helpful discussion.
14
The role of integrity in this context was first highlighted in B Williams ‘A Critique
of Utilitarianism’ in J J C Smart and B Williams Utilitarianism: For and Against
(Cambridge: CUP, 1973). Williams, it should be said, was not very clear about what
he meant by integrity or why it might be morally important.
9
does he have a duty to maintain his integrity rather than saving himself or others from
being harmed?
In investigating whether there is such a duty, let us further explore the selfregarding reason a person may have not to collaborate. We can then see whether the
person has a duty to do what she has reason to do. Many of us will feel that we have a
reason not to collaborate. We have a reason against making ourselves an Agent of an
evil. Following Frances Kamm,15 I capitalize Agent here deliberately to distinguish
being an Agent of another person from being a person who acts. All people are agents
– we all act. Not all people are Agents – we do not all act to further the intentions of
others.
Were D to give in to X’s threat D would make himself X’s Agent. D would
subordinate himself to X’s will, even when D could not share the end which X has set
as his own. D, it might be argued, has a powerful reason against forging this
relationship with X. In refusing to respond to X’s threats, D secures his independence
from X’s will. This comes at a price – the price of the threat being carried out.
Sometimes this price may be too high to make it permissible for D to secure his
independence in this way. Nevertheless, D secures his independence from X, which
has great value. Significant costs may be worth bearing in order to achieve this.
Could this self-regarding reason against collaborating with X be a selfregarding duty to do so? Perhaps. If D bends to X’s will too easily, D demeans
himself. He fails to recognise the significance of maintaining his independence from
X’s misdirected will. Perhaps self-respect requires (pro tanto) a person to execute the
intentions of others only if one can share, or at least tolerate, the ends that they have.
As D cannot tolerate the end that X sets himself, D has a duty not to act in service of
X’s will. I am not sure whether this is true, but let us suppose that it is.
This does not yet yield the verdict that the defence of duress ought not to be
made available to D. There are two reasons for this. The first is that the duty that I
have just referred is obviously defeasible. The question is how easily it is defeated.
Given that we are concerned with lethal threats, we may think that the self-regarding
15
See Intricate Ethics: Rights, Responsibilities, and Permissible Harm (Oxford: OUP,
2007) 312.
10
duty not to collaborate is insufficiently powerful to make a significant difference to
D’s overall obligations.
Secondly, the agent-centred argument does not establish that collaboration
violates anyone’s rights. Hence, although D may act wrongly in responding to X’s
threats, but he does not wrong V. But whether the duress defence ought to be
available intuitively depends not on whether D violates a self-regarding duty, but on
whether D violates V’s rights.
It is more difficult to justify publicly condemning and punishing a person for
violating self-regarding duties when compared with other-regarding duties. For this
reason, the agent-centred argument has at most a very limited role to play in
international criminal justice or in public social condemnation of wrongdoers. In
order to have a more profound effect on social and political morality, we must find a
victim-centred argument against collaboration. I find it much more difficult to
construct such an argument.
In bolstering this verdict, notice that D need not endorse X’s intention, or the
disrespect that X shows to V – the fact that he acts on the threat tends to suggest
otherwise. Perhaps it might be argued that D acts wrongly because he creates the
impression that he endorses X’s intention by acting to further it, and that creating this
impression is disrespectful. But D may not even publicly affirm that he endorses X’s
intention. It may be clear to everyone, including V, that D is only acting under duress.
That may have been true in Erdemovic itself. After all, Erdemovic tried to resist
shooting the victims, and may have publicly been threatened to participate in the
killings. Hence, it is at best unclear that collaborating with evil demonstrates the kind
of disrespect for the victim that can constrain D from killing V.
ii)
Two Principles of Deontology
A stronger reason to think that killing under duress is wrong is that killing under
duress violates stringent moral restrictions on killing some people to save others.
There are certainly cases where it is very clearly wrong to kill an innocent person to
save one’s own life, so clearly wrong that the person who does so could never be
excused. Consider:
11
Human Shield: X is attempting to shoot D. D grabs V, an innocent bystander,
and uses V as a human shield. X shoots V several times killing her.
It is clearly wrong for D to use V as a human shield to prevent himself from being
killed. It is so clearly wrong to do this, it might be argued, that D ought never to be
excused for killing V. D, along with X, murders V.
Here is an argument why it is very clearly wrong for D to kill V to save his
own life: D violates two well-established moral doctrines. First, D violates the
Doctrine of Doing and Allowing (DDA). It is more difficult to justify killing a person
than refraining to save a person. In order to justify killing a person it must be shown
that doing so averts a much more significant threat than will be realized if he fails to
kill. In Human Shield, the threat that D will face if he does not kill V is no greater
than the threat that V faces if D kills V by using him as a human shield. Hence, it is
impermissible for D to kill V.
Secondly, D violates the means principle.16 If D uses V as a human shield, D
uses V as a means to save his own life. It is often regarded as especially difficult to
demonstrate that it is permissible to use a person as a means to serve one’s own end.
We are permitted to use a person as a means to an end in a way that harms her to
some degree, n, only if that person has an enforceable duty to serve that end even if
she would be harmed to degree n. V lacks an enforceable duty to save D’s life at the
expense of V’s life. As V would not be expected to give his life up for the sake of D’s
life, it would be wrong for D to use V as a means to save his own life.
Obviously, both the Doctrine of Doing and Allowing and the means principle
are highly controversial. I will not explore the foundations of these principles, or their
best articulation, here. Each principle, even if controversial, helps support the intuitive
verdict that D has no defence for killing V in Human Shield.
Now, if each of these principles – the Doctrine of Doing and Allowing and the
means principle – is valid, we would seem to have an argument in favour of the
position of English law, and following it the ICTY, on the restriction on the
availability of the duress defence. For in standard cases of duress, it is also true that
the DDA and the means principle are implicated.
16
For reasons beyond the scope of this paper, I think it better to refer to the means
principle here than the more familiar Doctrine of Double Effect.
12
Consider:
Standard Duress: X threatens D that if D does not kill V, X will kill D.
If Standard Duress is no different to Human Shield, we might doubt the verdict that
some endorse that D would have a defence (either justification or excuse) to murder.
After all, D should not be justified or excused in Human Shield.
And Standard Duress shares morally salient features with Human Shield. If D
kills V, D violates the DDA. D kills rather than allowing V to die. Furthermore, if D
kills V, D uses V as a means to save his own life. The killing of V is the means by
which D influences X’s intentions, averting the threat that he faces from X. This, it
might be argued, is no different from using V as a means to save his own life, in cases
such as Human Shield.17
In what follows, I will suggest that whilst these arguments are indeed
powerful, there are countervailing arguments that militate in favour of permitting the
duress defence to killing in at least in some cases. This is true for at least three
reasons. First, it is not always true that a person who kills under duress kills a person
as a means to an end. In such cases, whilst the DDA applies, the means principle does
not. Killing may be justified, in that case, if the threat averted is significantly greater
than the harm caused. Secondly, it makes a significant difference to the permissibility
17
Jeff McMahan picks out these two features of morality as reasons to doubt the
validity of the duress defence as a justification or a complete excuse to killing in war.
About the standard case of duress where X threatens you with death unless you kill V,
he writes: ‘First, to accede to the demand of the person who threatens you would
involve killing, whereas to refuse would involve allowing someone – yourself – to be
killed. And it is, and always has been, widely accepted that the moral presumption
against killing is in general stronger than the presumption against allowing someone
to be killed (that is, failing to prevent someone from being killed). Second, killing the
innocent third party would involve intending a harmful death, while refusing to kill
him would not. And it is, and always has been, widely accepted that the moral
presumption against bringing about a harmful death as an intended effect is in general
stronger than the presumption against brining about a harmful death as a foreseen but
unintended effect.’ Killing in War 113.
13
of killing a person that the person is already doomed to die. This factor was rightly
considered important by Judge Cassese in Erdemovic, even if it was not clear that it
was decisive in that case. Even killing a person as a means to an end may be justified
where that person is already doomed to die. Thirdly, there are important differences
between Standard Duress and Human Shield. Although the person killed is used as a
means, she is not killed opportunistically, in a sense to be explained, and this makes a
significant difference to the permission to kill her.
III.
Duress and Side-Effect Killing
Killing as a side-effect rather than intentionally is sometimes murder. Consider:
Insurance. D plants a bomb in D’s shop. He does this to destroy the shop for
the insurance money. He is certain, when he detonates the bomb, that the
explosion will kill V, a passerby.
D foresees V’s death as a certain side-effect of his action rather than intending it. In
English law D murders V, and this seems the right verdict.18
Cases like this suggest that if D kills V as a result of duress, V is not
necessarily used as a means. The reason for this has to do with the structure of
intentions involved in duress cases. The X, in making the threat, intends that D act in
a certain way. D intends to act in that way in virtue of the fact that if D executes X’s
intention, X will not carry out the threat. Now suppose that X threatens D that he will
w unless D does an act which will certainly kill V. Whether V’s death is the means by
which X’s threat is averted depends on whether averting the threat depends on V’s
death. If X does not intend V to be killed, D need not intend to kill V either. Killing V
is thus not the means by which X’s intention is fulfilled. Hence V is not used as a
means to avert the threat posed by X.
To illuminate this possibility consider:
18
See R v Woollin. Wollin has the distracting feature that side-effect killings are
included within the definition of intention. In principle, the jury has the latitude not to
call this a murder. They would almost certainly convict in this case though.
14
Insurance Threat: X wants his shop blown up for the insurance money. He
threatens D that if D does not detonate a bomb that X has planted in X’s shop,
X will kill D and four other members of D’s family. D is certain, when he
detonates the bomb, that the explosion will kill V, a passerby.
The killing of V is a side-effect of the execution of D’s intention to blow up the shop.
Furthermore, the killing of V is not the means by which X’s threat to D and his family
is averted. Hence, although the DDA is implicated, the means principle is not.
Does the DDA render it wrong for D to kill V? The DDA is normally thought
less stringent than the restriction on killing a person as a means. It is at least
sometimes permissible to kill a person as a side-effect to avert a lethal threat to more
people. For example, it is permissible to turn a trolley away from five people towards
one who will certainly be killed.
However, it may not always be permissible to kill one person as a side-effect
of saving five. Consider Philippa Foot’s example: an antidote could be created that
will save five lives but creating it will also give rise to lethal fumes that will kill one
person. It is to some extent intuitive that it is wrong to create the antidote even though
the killing of the one person is a side-effect of the act which saves the five.
I am unsure about the relevant differences between these cases. Perhaps it is
that the person creating the antidote will create a new threat in Foot’s case, but will
only divert an extant threat in the trolley case. But it is also not always true that it is
wrong to create a new threat as a side-effect of one’s act of saving others. For
example, in the context of a just war, it is intuitively permissible to bomb munitions
factories even if civilians will be killed as a side-effect. But doing so creates a new
threat rather than merely diverting a threat.
It is contested ground, then, whether and when it is permissible to harm a
person as a side-effect of the saving of others. I find it intuitive that D is permitted to
blow up the shop to save himself and his family in Insurance Threat. This may be in
part because the threat that D averts is to himself and his family, in contrast with
Foot’s case or the trolley case. It may be for other reasons to do with the structure of
duress cases that I will explore below.
Now, in the context of war, it is at least sometimes true that the pro tanto
wrongful killing that a soldier does under duress is side-effect killing rather than
killing as a means. It may be pro tanto wrong for a soldier to go to war, or to commit
15
a particular act of war, not because the soldiers that he kills would wrongfully be
killed, but because the side-effect harms that will be caused by the war will be
disproportionate. Even if duress is not a defence to engaging in a war that is unjust in
virtue of the fact that it lacks a sufficient just cause, it may sometimes be a defence to
engaging in a war that has a just cause but that is disproportionate in its side-effects.
Perhaps it might be argued that duress is never a defence for wrongfully
engaging in a disproportionate war in virtue of the fact that the threat will never be
sufficiently grave to outweigh the deaths that a soldier will cause as a side-effect.
Against this, it is important to remember that a soldier entering a war will often be
uncertain about whether he will cause any deaths as a side-effect of his action.
Suppose that a soldier is threatened that if he does not engage in a
disproportionate war he will spend the rest of his life in jail. He goes to war and kills a
civilian as a side-effect. Let us also suppose that, were it not for the threat that the
soldier faced, the risk to the civilian would have rendered the act of war
disproportionate. It follows that were it not for the threat that the soldier faced, the
soldier would have committed an unjustified homicide – the equivalent of what, in
English law, is called manslaughter. Nevertheless, the soldier may have been
permitted, all things considered, to engage in the act of war. The fact that he faced a
threat of imprisonment for life that would be averted only if he engaged in the act of
war may render it permissible to impose what would otherwise be a disproportionate
risk of harm on the civilian. Duress, in this case, may provide a defence to
manslaughter.
IV.
Killing Those who are Doomed to Die
One feature of Erdemovic that rightly moved Judge Cassese, though not the majority,
was the fact that those that Erdemovic killed may well have been doomed to die
whatever he did. Unlike Cassese,19 I doubt that this is the only case where the
proportionality criteria is met in a case of killing under duress. It is nevertheless
clearly a morally significant fact about Erdemovic.
i)
19
When is Pareto Decisive?
See para.12.
16
In standard duress cases, as I have already suggested, D kills V as a means to avert the
threat that X would otherwise pose to D and others. In Erdemovic, the killings have
this feature – Erdemovic killed his victims in order to affect the decision of his
superiors so that they would not kill him. The question is: is killing a person as a
means permissible if that person is already doomed to die?
This problem is familiar from Bernard Williams famous Jim and the Indians
case. Here is a summarized version:
Jim and the Indians. Jim finds himself in the central square of a small South
American town. The Captain has tied twenty Indians against a wall. He is
about to kill them. The Indians have been randomly selected to remind people
of the disadvantages of protesting. As Jim is an honoured visitor, the Captain
makes him an offer that if he kills one Indian from the group then, as a special
mark of the occasion, the other Indians will be set free. Otherwise all twenty
will be killed. There is no other way for Jim to save anyone. The Indians and
other villagers beg Jim to accept.20
Most people think that Jim is at least permitted to kill one Indian. Many think that he
is required to do so.
Central to explaining these judgements is this. Compare
1) Jim selects and kills one Indian; with
2) Jim does nothing.
Jim selecting and killing one Indian pre-empts the Captain killing 20. Hence, 1)
renders no Indian worse off overall than 2). 1) renders 19 Indians much better off than
2). Many people conclude that as long as the Indian killed is selected fairly, Jim does
nothing wrong all things considered in killing an Indian - it is Pareto optimal for Jim
to kill an Indian.
20
‘A Critique of Utilitarianism’ 98-9.
17
Nevertheless, as Williams suggested, Jim and the Indians is a troubling case.
Williams thought it a failing of utilitarianism, and other versions of consequentialism,
that it could not explain why the case is troubling. For consequentialists killing the
one Indian is pure benefit. But, Williams thought, Jim’s conduct is troubling because
killing an Indian would damage Jim’s integrity. This, Williams thought, was closely
related to the fact that Jim himself would do the killing. But, Williams believed, it
was not simply because Jim would be killing rather than letting die that explained our
uneasiness. It is rather the fact that in killing there would be a conflict between Jim’s
actions and the projects that he is committed to and that he values. 21 Whilst, for
Williams, this did not necessarily render the conduct wrong, it should at least give us
pause for thought. Consequentialists don’t pause.
For reasons outlined above, I doubt that integrity is very important in an
investigation into whether it is wrong for Jim to act. Williams claim that we should
feel uneasy about the permissibility of Jim killing one of the Indians can be vindicated
though. I doubt that all Pareto optimal acts are permissible, and this seems true in
variations on the cases that we are considering. For example, suppose that there is
only one Indian who is doomed to die. Jim can kill this Indian or let the Captain do
so. If Jim kills the Indian, he will receive a reward of some jellybeans that will
otherwise go to waste. It is Pareto optimal for Jim to kill the Indian – the Indian is no
worse off than he would be were the Captain to kill him, and Jim is better off in virtue
of having received jellybeans. No one else is affected. Yet it is surely wrong for Jim
to kill in this case.
Now consider the following:
Duress (Doomed Variation): X threatens D that if D does not kill V, X will
kill D. X will also immediately kill V.
Doomed Variation is similar to Jim and the Indians, except that it is D’s life that is
threatened rather than nineteen innocent strangers. I believe that it is permissible for
D to kill V in this case.
One way to reinforce this verdict is to consider V’s perspective more closely.
V will be certainly be killed: either by D or by X. Suppose that V would prefer to be
21
‘A Critique of Utilitarianism’ 116-7.
18
killed by X than D. Suppose also that V is able to select his killer. If V selects D as
his killer, D’s life will be saved. V’s preferences, in this case, do not seem decisive. D
is required to select D as his killer given that this will save D’s life. There would be
some restriction on V’s options were he not permitted to select X as his killer. Given
that V’s autonomy is already very severely constrained – he will certainly be killed
almost immediately – we have a powerful reason to act on his preferences if we can.
But in this case D is not required to act on V’s preferences to be killed by X even if V
has, and expresses, those preferences. This is because of the very great importance of
D’s life.
Hence, if D kills V, D acts in a way that V would be required to authorize him
to act were he able to do so, given the constrained set of options available to D. If D
kills V, D shows sufficient respect for V, for D acts in a way that V would have a duty
to authorize. I do not say, I should emphasize, that a hypothetical duty to authorize is
always required to render killing permissible. Furthermore, if actual authorization can
be secured, D ought to attempt to secure it. It respects another person more to treat
their authorization as important even when it is not required. Compare taking, rather
than asking for, another person’s fire hydrant when it is required to put out a fire.
But in the absence of actual authorization, the fact that V would have a duty to
authorize D to kill him provides a powerful reason in favour of permitting D to kill V.
This is so even if D, rather than others, is the beneficiary of the killing. Even if D
killing V is not what V wills, it is what V has a duty to will. In this way, D’s killing of
V would not show disrespect to V as an independent reasoning agent.
ii)
Self-Selection and Collaboration
This kind of argument is unavailable in a related case. Consider:
Choose Oneself: X threatens a group of twenty people, who are identical in all
morally salient respects, that he will kill them all unless any one of them kills
the other nineteen. The one that kills the other nineteen will then be saved.
Each person is separated from the other nineteen and there is no way in which
they can communicate with each other. Each can kill the other nineteen simply
by pressing a button. The person who presses fastest will survive.
19
X has not picked out any member of the twenty for saving. Any person can save
herself by killing the others immediately. Let us suppose that D, a member of the
twenty, kills the other nineteen. He is then saved. Does D act wrongly?
This case is important to evaluate, in part because it may play an important
role in evaluating the role of collaborators in Nazi concentration camps. Primo Levi
suggests that collaborating improved one’s chances of surviving the camps. Here is
his explanation:
The privileged prisoners were a minority within the Lager population, but they
represent a potent majority among survivors; in fact, even if one does not take
into account the hard labour, the beatings, the cold, the illnesses, it must be
remembered that the food ration was decisively insufficient even for the most
frugal prisoner: the physiological reserves of the organism being consumed in
two or three months, death by hunger, or by diseases induced by hunger, was
the prisoner’s normal destiny. This could be avoided only with additional
food, and to obtain it a large or small privilege was necessary; in other words,
a way, granted or conquered, astute or violent, licit or illicit, to lift oneself
above the norm.22
The real world case that Levi describes many empirical and normative complications
that Choose Oneself lacks. It is not clear whether more people survived the camps as a
result of collaboration. Perhaps collaborators each increased the number of deaths
overall. It is also concerned with depriving others of resources rather than killing
others in a more direct way. Finally, it is not clear that the same people died as a
result of collaboration who would have died were collaboration not to have occurred.
Choose Oneself is cleaner - it raises fewer moral considerations. Nevertheless,
there are similarities with Levi’s analysis of the Lager and Choose Oneself. In the
Lager, there was presumably no real prospect of fairly selecting a person for
collaboration. A person must select herself to be ‘above the norm’. If no one does
this, all will die. But doing so does render others worse off – it deprives each person
of an opportunity to select herself to be the person who is ‘above the norm’, and thus
to save herself.
22
The Drowned and the Saved 26.
20
It might be permissible for any person to press the button as quickly as
possible in Choose Oneself. To see why, let us begin by altering one feature of the
case. Suppose that the twenty could find a fair procedure for selecting the person who
will kill the other nineteen. Each person would have a strong reason to participate in
such a procedure. Each person would be better off ex ante than she would be were no
one selected. For each person would have better prospects as a result: a one in twenty
chance of survival as opposed to zero chance of survival. Furthermore, no one will be
worse off ex post than she would be were no one selected. Each person who is killed
would immediately have been killed anyway. Finally, one person would be much
better off ex post than she would be were no one selected. These considerations
militate very strongly in favour of permitting the twenty to operate a fair procedure
for selecting one killer.
In such circumstances, it seems wrong for any person to press the button
without setting up a fair procedure. It may be argued that the procedure whereby the
fastest to press survives is itself fair. I doubt it. It advantages those who are
unscrupulous, who reason quickly, who press quickly, and so on. Perhaps it might be
argued that it is a matter of luck whether one has these characteristics, and so the
procedure that relies on them, being random, is itself fair. This seems difficult to
defend for the following reason. If a person is entitled to benefit himself, in
procedures like this by relying on their characteristics which tend towards quick
pressing, the slow, indecisive and conscientious will systematically lose out to the
fast, decisive and unscrupulous.
To see this, imagine a series of procedures where the first one to press a button
receives a benefit. If a person is permitted to press as quickly as possible, the fast,
decisive and unscrupulous will receive many benefits over the course of their lives
when compared with the slow, indecisive and conscientious. We can expect that
people with these characteristics will have secured many other advantages for
themselves over the course of their own lives already. For this reason, we should
prefer procedures that do not rely advantageous characteristics that people have.
Pressing the button as quickly as possible is, in this way, unfair.
Where the twenty cannot communicate with each other, though, there is no
prospect of a mutually agreed fair procedure. Nevertheless, each person could try to
act in accordance with what fairness requires. One possibility is that each person
could put his name in a hat with nineteen blank pieces of paper. Only if his name
21
comes up will he press the button. Otherwise he refrains.23 This suggestion has the
following deficiency – it is possible that no one amongst the twenty selects the piece
of paper with his own name on it. If this occurs, all twenty will be killed. We have
strong reasons to exclude this possibility.
Another possible solution is that each person should press the button only at
the moment when, from the evidence that he has, he will have a one in twenty chance
of being saved. That would give him the chance of being saved that he would have
under a fair procedure. If he does this, it might be argued, he respects the other
nineteen by refraining from taking for himself more than his fair chance of survival.
This solution is also problematic. To see this, imagine that D, one of the
twenty, knows that the other nineteen are committed pacifists. They will definitely
refrain from pressing the button themselves. In that case, at any moment that D
presses the button, his probability of surviving, from his evidential perspective, is one.
If D is not permitted to press the button until his prospects of survival are one in
twenty, he is never permitted to press the button, and all twenty will be killed. I very
much doubt that the fact that the other nineteen people are pacifists can render it
wrong for D to kill them to save his own life.
A better solution is that each person must act in a way that will give an equal
chance of survival to those who wish to have such of survival. If there are ten
committed pacifists and ten who want an equal chance of survival, D should press the
button at the moment when he has a one in ten chance of surviving. This is the
maximum chance that D can have of survival consistent with providing an equal
chance to those who want it.
Similar considerations will be important in the circumstances that Levi
describes above. Suppose that collaboration by a particular person does not increase
the number of deaths overall. However, by collaborating, a person may deny others a
chance to collaborate. If one person takes the position of an administrator in a
concentration camp, that position is not open to others. By taking this position, this
person may disadvantage the others by denying them an opportunity to collaborate.
Each person ought, in that case, to provide an equal opportunity of collaboration to
those others that want to collaborate.
23
I thank Matthew Clayton for this suggestion. I should add that he does not endorse
it.
22
Even if collaboration did not increase the number of deaths in the camp, then,
many collaborators will have acted wrongly. This may not have been in virtue of the
fact that collaboration caused more deaths overall than would have occurred without
collaboration – that will have been true only in some cases, and it will often be
difficult to establish. It may have been true in virtue of the fact that those who
collaborate raised themselves above the norm in an unfair way. Had they done so
fairly, collaboration might sometimes have been permitted.
V.
Manipulation, Opportunism and Duress
In standard cases of duress, the arguments developed in the previous section do not
apply. For example, in Standard Duress, V is killed as a means to avert the threat that
X will otherwise pose to D, V is not doomed to die, and it is much worse for V if D
kills than if D does not kill. Standard Duress may seem no different to Human Shield
– both the DDA and the means principle are implicated in both cases.
Yet, whilst killing under duress does seem difficult to justify, and may be
unjustified in Standard Duress, it is unintuitive that Standard Duress is on a par with
Human Shield. Suppose that Standard Duress is revised so that the threat that X poses
to D is not only to D but also to four members of D’s family. I find it intuitive that D
may kill V. Yet using an innocent person as a human shield seems wrong even to
protect oneself and four members of one’s family. Here I suggest two important
differences between these cases.
The first difference, is that as D would be acting as X’s Agent. Drawing on
Kamm,24 it might be argued that as X would be fully responsible for the killing, D’s
responsibility for killing is weakened. This can help to justify D killing V. Kamm
herself denies that it is permissible to kill one to save five in cases of this kind.25
Given her argument, I am unsure whether Kamm is right.
24
Intricate Ethics ch.10. See, also, The Moral Target: Aiming at Right Conduct in
War and Other Conflicts (Oxford: OUP, 2012) ch.5.
25
Intricate Ethics 306. Kamm treats this case as similar in important respects to using
another person as a means to the good. See 337, n.7. ‘Redirecting’ X might, though,
be permitted. See 307.
23
To appreciate the second difference, we need to grasp a distinction that
Warren Quinn made in his influential work on intentional killing that is not widely
noticed. It is the distinction between using a person as a means to an end
(manipulative agency) and harming a person opportunistically (opportunistic agency).
These two ideas are often treated as identical and collected under the label
‘opportunistic agency’. They are better distinguished. Standard Duress involves
manipulative agency but not opportunistic agency. Human Shield involves both
manipulative and opportunistic agency. Standard Duress thus lacks one morally
salient feature of Human Shield, rendering it easier to justify killing in Standard
Duress than in Human Shield.
i)
Agency and Responsibility
In duress cases, D acts as X’s Agent. Although D executes an intention to kill V, he
does so due to X’s command. Ultimately, is X’s intention that D is executing, it might
be said, rather than D’s. Kamm argues that in these circumstances, X has full
responsibility for the killing. In virtue of this fact, D lacks some degree of
responsibility. This is so, she suggests, partly for a complex retributivist reason. If D
kills V, X will deserve to suffer for his crimes. When D is threatened by X, there is a
special reason for D to kill V: if D kills V, X will then deserve to suffer. As X has
threatened D, Kamm suggests, D deserves to deserve to suffer. There is what we
might call a ‘meta-retributivist’ reason for D to kill V.
I very much doubt that this is the right explanation of the permission for D to
kill V in duress cases. One reason is that we should doubt retributivism in general.26
Another is that if retributivism is true, surely we have a standing reason to ensure that
people do not deserve to suffer. Suppose that X attempts to kill Y. Even if
retributivism is true, I surely owe it not only to Y, but also to X, to prevent the killing
if I can do so at little cost. In that way, I could prevent X deserving to suffer.
Perhaps it might be argued that this is a reason for D to kill V. If D kills V, X
deserves to suffer for one death. If D refrains from killing V, X will kill five. Hence,
D ought to kill V. He ought to do so to reduce X’s amount of deserved suffering.
Again, given doubts about desert, I suspect that we should reject this view. Even if
26
See, further, V Tadros The Ends of Harm ch.4.
24
there is such a thing as desert, I doubt that it provides a powerful reason for D to kill
V.
A better approach focuses more closely on the nature of agency. Consider
more carefully the relationship between X’s intention and the intention which D
forms and executes. D forms and executes the intention to kill V only because X has
the intention that D forms and executes the intention to kill V. The content of D’s
intention, in this case, depends almost completely on the content of X’s intention. The
intention which D forms is simply the intention that X wishes him to form. D would
form and execute any intention that X wished him to form and execute. Or at least this
is true within some distant limits – D would not kill all of his family and others in
virtue of the threat that X poses to him and his family.
Here is another way to understand this. D’s first order intention – to kill V depends on a second order intention that D has formed and executed – namely to form
and execute the intention that X intends D to form and execute. This idea is neatly
captured in cases where X first poses the threat to D and his family, and D responds
by saying ‘I’ll do whatever you want’. Forming a second order intention to form and
execute the intentions that another person wants one to form and execute is a central
component of becoming another person’s Agent. It is not the only component of
becoming another person’s agent. The other person, let us call that person ‘the
director’, must also intend to issue instructions to their agent. Agency involves a
relationship between an agent and director where the director intends to issue
instructions to the Agent about what to do and the Agent intends to do what he is
instructed to do.
Now, it might seem that whether one is an Agent can make no difference to
what one is permitted to do. This might be thought true on the following grounds. D
needs to decide whether to act as X’s Agent. In deciding whether to do this, D must
consider what he will be instructed to do. If D will be instructed to act wrongly, D
ought not to become X’s Agent. If X instructs D to act wrongly, D’s act cannot
become permissible in virtue of the fact that D acts as X’s Agent. For any
consideration that renders acting in this way wrong were D not X’s Agent would
count in exactly the same way against D becoming X’s Agent.
For example, if it is wrong for D to kill V as a means to save five were D not
X’s Agent, D also has a decisive reason not to become X’s Agent. For were D to
become X’s Agent, D would form an intention to follow X’s instruction to kill V as a
25
means to save the five. As it is wrong for D to do this, it is wrong for D to become
X’s Agent.
I begin by showing that this argument at least sometimes fails. To see this, it is
crucial to notice that a director sometimes has a right to do wrong. It is at least
sometimes the case that if a director instructs his Agent to do something that it would
be wrong for the director to do, the Agent does not act wrongly. He may not act
wrongly if the director had a right to do wrong, and the reasons that militate in favour
of him having such a right also militate him being permitted to seek help to act
wrongly.
Here is an example:
Rock and Roll. X wishes to play loud rock music on her stereo. However, she
suffers from a physical disability which prevents her bending down to turn her
stereo on without suffering great pain. D is X’s home help. X instructs D to
turn the music on loudly. This disturbs V, X’s neighbour. X is unmoved by
this.
Let us suppose that would be wrong for X to put her music on this loudly. However,
let us also suppose that it would be wrong for other people to interfere to prevent her
from doing this – there is good reason for X to have a liberty right to play loud music,
even when this wrongs V. In that case, it may be permissible for D to help X to
execute her intention of playing loud music by turning the loud music on.
The reason is that we have a powerful reason to ameliorate the extent to which
X’s liberty has been restricted as a result of her disability. We can do this by allowing
X to have D act as her agent. X’s liberty is enhanced in virtue of the fact that D forms
the second order intention to do what X wishes her to do. Because D has formed this
intention, X can do more – she has a tool that is responsive to her instructions. If X
instructs D to put the music on, X acts wrongly. D, though, does not act wrongly. It is
only X and not D that is morally responsible for the wrongful disturbance caused to
V.
This analysis of Rock and Roll demonstrates that it is sometimes permissible
for an Agent to carry out an instruction that it was wrong for the director to issue.
Furthermore, it demonstrates that this is sometimes true in virtue of the fact that
Agency eliminates responsibility for wrongdoing. It is at least sometimes true that the
26
director is solely morally responsible for the wrongful act that the Agent helps her to
perform in virtue of the fact that the Agent is acting as an Agent.
However, it also provides limited support to the importance of Agency in
duress cases such as those we have been considering. The reason is that this analysis
of Rock and Roll depends on the idea that there is a good reason for D to ameliorate
the restriction on X’s liberty that arises in virtue of her disability. This is not a feature
of cases such as Standard Duress. In Standard Duress D has no reason to enhance
X’s liberty to kill V by deciding to act as X’s Agent. All that we have shown, then, is
that there are some cases where moral responsibility for wrongdoing is reduced as a
result of Agency. We have not shown that this idea is relevant to duress cases.
We need a different argument why Agency can sometimes make a difference
to responsibility. It will help us to understand the issues better if we consider first the
decision to form the second order intention to follow the instructions of the director.
Consider:
Inchoate Duress: X sincerely and credibly threatens D that if D does not do
what X instructs D to do X will kill D’s child.
D has a good reason to form a conditional intention to do what X instructs him to do.
Within some limits – wide limits – D is required to do as X tells him. For example, if
X tells D to tie X’s shoelaces, or to go to the shops for milk, etc etc, D ought to do
these things. There are some things, though, that D clearly ought not to do if X
instructs him. He ought not to kill two of his children to save the one child threatened
by X, for example. This is a condition on the second order intention that D ought to
form in the face of X’s threat. X should think: I will do some range of actions (let us
call them ‘permitted actions’) if instructed to do them but I will not do some other
range of actions (let us call them ‘prohibited actions’) if instructed to do them.
We can now restate the question of duress in the following way. In duress X
threatens D to become X’s Agent. If D’s threat is sincere, credible and serious, D has
good reason to become X’s Agent. The question is: what is the range of prohibited
actions that ought to condition D’s intention to act as X’s Agent?
In order to answer this question, we might think about the circumstances in
which we would be required not to provide another person with a tool that will be
used for killing. In forming the intention to act on another person’s instructions, one
27
makes oneself into a tool for that other person. If one can provide a tool for another
person in order to avert a threat, perhaps one can also make oneself into a tool for
another person in order to avert that threat. But providing another person with a tool
does not seem stringently controlled by the means principle. To see this, consider:
Sword: X wishes to kill V with a sword. Only D has access to the sword. If D
does not provide X with the sword, X will kill D and four members of his
family.
Even though D ‘does’ rather than ‘allows’ in this case, D’s act seems to me
permissible. D, in this case, does not use V as a means to avert the threat. He provides
X with a tool to kill V, but he does not use V himself.
Perhaps it makes a difference, in duress cases, that the person doing the killing
is herself the tool that is used to do the killing. I doubt that this feature is terribly
significant, though. Compare:
Human Club: X wishes to kill V. He can do this by using D as a human club,
but only if D holds his body rigid. X will swing D at V, killing him.
D, in this case does not simply provide X with a tool, he makes himself X’s tool.
Nevertheless, this seems to me permitted. In deciding to hold his body rigid, D does
not use V as a means to save the five in the objectionable way that V is used in cases
such as Human Shield.
Making oneself another person’s agent is somewhat different to making
oneself into a human club, though. In making oneself into an Agent, one forms an
intention to form and execute the intentions one is instructed to form. When these
intentions are carried out, the harm that one imposes on others is also imposed
intentionally (or at least, one has the intention to affect others). In making oneself into
a human club, in contrast, one simply intends to turn oneself into a tool for others to
use.
Nevertheless, perhaps the conditions under which one may make oneself the
Agent of another are not terribly dissimilar to the conditions under which one may
make oneself into a tool for the use by another. When one acts as an Agent, one treats
oneself as a tool for the other’s use. This negates, to some extent, the responsibility
28
that one has for one’s own actions. Making oneself into a tool for another’s use, by
forming the intention to follow another’s instructions, may be permitted in the same
circumstances as providing another person with a tool to further their own ends is
permitted. If D is entitled to provide X with a sword in Sword, D is also permitted to
kill V if that is the only way to avert the threat to himself and four members of his
family.
This argument seems to me to have some force. In making oneself another
person’s Agent, one treats the first-order content of one’s intentions as less important
in guiding one’s actions. It is the second-order intention – the intention to form and
execute the intentions one is instructed to form and execute – that guide one’s actions.
One can sometimes have good reason to give priority to such a second-order
intention, as we can already see from cases such as Rock and Roll. In duress cases,
one has a similar reason to give one’s second order intentions priority in guiding
one’s actions – doing so is necessary to avert a significant threat. In doing so, one
relegates the significance of one’s first order intentions in one’s decisions. This
distancing can ground a denial that one is responsible, or at least as fully responsible,
for the content of those intentions.27
In short, D can deny that he uses X as a means. He may claim, on the contrary,
that he has allowed himself to be used by X as a means. Whether he is wrong to do
this depends primarily on whether the threat that X imposes is sufficiently significant
to render it permissible for D to provide X with a tool to serve his purposes, and that
does not depend on showing that it would be permissible for D to use V as a means to
avert X’s threat. I am not sure that this argument is decisive, but it seems to me to
have some force.
ii)
Opportunism and Manipulation
There is another important difference between Standard Duress and Human Shield
that should incline us to relax the availability of the defence of duress. Let me begin
27
This also helps to support Kamm’s view, defended in Intricate Ethics ch.10, that it
makes a difference whether the person acting is responding to an offer from the threat
maker, or whether she instigates the offer. It does so, though, on different grounds to
those she offers.
29
by crudely describing a difference between these cases. In Human Shield, V is simply
a bystander – the threat that is posed to the five has nothing at all to do with V. If V is
used as a means he is compelled to be involved in a scenario that he otherwise has no
involvement in. In Standard Duress, in contrast, V is already involved in the threat
that is posed to the five. The threat to the five comes about in virtue of the fact that X
has targeted V.
In order to investigate the difference this might make, it will be helpful to
attend to a distinction that often goes unnoticed in Quinn’s work on Double Effect –
the distinction between using a person as a means and opportunistic agency. Often,
scholars referring to Quinn treat these different ideas as though they are
interchangeable. However, Quinn intended to pick out two different features of cases
like Human Shield. One feature is that if D kills V, D uses V as a means to save the
five. To put it another way, if D kills V, V is D’s tool. Let us say that if D kills V, D
manipulates V.
The other feature is that if D kills V, V will have provided D with an
opportunity to save the five that D would have lacked were it not for V’s presence.
Hence, if D kills V the five will have benefited from V’s existence, or presence at the
scene. In contrast, if D does not kill V, the five are no worse off than they would have
been had V not existed, or was not present at the scene. Let us say that if D kills V, D
acts opportunistically – he exploits an opportunity created by V’s presence.
Furthermore, V’s existence or presence makes the five no worse off. V makes no
causal contribution to the threat that the five face. Hence, if D kills V, he does not
eliminate a threat that V makes a causal contribution to. D, we can say, does not act
eliminatively.
Each of these features of Human Shield seems to have some normative
significance. First, a person can object to being used as a tool to serve an end. It is
especially wrong, it might be argued, to use a person to serve one’s ends. Each person
is entitled to set ends for herself. She may not have her ends imposed on her by
others. Using a person as a means is a way of imposing ends on a person. It treats her
as though she exists to serve these ends. This is inconsistent with her status as an endsetter.
Second, a person has a more powerful complaint against being harmed if she
has made no contribution to the threat that others face. If harming a person merely
neutralizes a threat that she poses, harming her is easier to justify. A person can often
30
be expected to internalize the costs of her own existence or presence. She should aim
to ensure that her existence or presence is not harmful. But a person is not normally
required to benefit others. She need not ensure that others profit from her existence or
presence. Hence, harming a person opportunistically is more difficult to justify than
harming a person eliminatively.
Obviously, there is much more to say about these ideas than I have space for
here. I do not intend the sketch that I have provided to be a defence of them. I have
only sketched an account that I hope reveals that they have some intuitive force. I
hope also to have revealed that the moral idea grounding the restriction on
manipulative agency is plausibly independent of the moral idea grounding the
restriction on opportunistic agency.
With this distinction in hand, we are in a better place to evaluate Standard
Duress. In Standard Duress it is true that if D kills V, D uses V as a means. The
killing of D is the means by which X is motivated not to carry out his threat to kill D.
Hence, the killing of V is best seen as manipulative. If I am right in my
characterization of what is wrong with manipulative agency, we can articulate an
objection that V might make to this: V might claim that he has been used to serve the
end of motivating X not to kill D at the cost of his life. This, he might argue, is not an
end that he is required to serve. He has been used to serve D’s end of averting a threat
posed by X.
But if D kills V in Standard Duress D does not act opportunistically. The
reason for this is that it is the fact that X wants V dead that gives rise to the threat that
X poses to D. If V had not existed, or had not been present, X would not have posed a
threat (or at least this threat) to D. If D kills V, D eliminates a threat that exists in
virtue of V’s existence or presence. D might thus claim that if he is not permitted to
kill V, V’s existence or presence will have rendered D worse off. V, D might claim,
can be expected to internalize the costs of his own existence. This will be done if D
kills V.
Of course, it is also true that if D kills V, V is disadvantaged by D’s existence.
This suggests that the fact that killing of V is eliminative rather than opportunistic
cannot alone be decisive in grounding D’s permission to kill V. My aim, though, has
only been to point to an important moral difference between Standard Duress and
Human Shield that tends to go unnoticed, and that might have a role in weakening the
constraint on killing under duress. Unlike most cases where one person will be
31
harmed as a means to avert a threat, such as Human Shield, killing under duress is not
typically opportunistic. The person doing the killing does not benefit from the
existence or presence of the person killed.
Conclusion
There are many other issues in duress cases that I have not considered. My aim has
been to show that there are circumstances in which duress can justify killing a person
where that killing would otherwise be wrong. Furthermore, I have aimed to show that
the restriction on harming a person under duress is not as strict as the restriction on
harming others as a means. These results seem important in helping to support
Cassese’s dissent in Erdemovic. A blanket restriction on duress as a defence to
homicide, even to crimes against humanity, seems to me unwarranted, at least if the
law aims to track morality reasonably closely.
32