Like many other philosophers, I reject consequentialism in
favour of a more deontological approach to ethics. That is, I favour a moral theory
that implies that in certain cases, one ought to refrain from harming people in
certain ways – even if, from an impartial agent-neutral perspective, the
world as a whole would be a better place if one did harm those people in those
ways. The notorious “trolley problem” is one good way to try to understand just
what principles underlie these agent-relative, deontological restrictions:
Five people are trapped on the railway track, and a runaway
railway trolley is hurtling towards them. In this case, it seems permissible to
divert the trolley away from the main track (where it will crush five people), onto
a side track where it will crush only one person. On the other hand, it does not seem permissible to push an innocent
bystander onto the railway track so that after smashing into him, the trolley will grind
to a halt before it can reach the five people who are trapped further down the
In addition to this classic case, there are many other cases of
this general sort, which have been extensively discussed by such philosophers
as Philippa Foot, Judith Thomson, and Frances Kamm, among others.
In each of these cases, there are two options: either to
divert the trolley or to refrain from doing so. Which of these options is the right
thing to do obviously depends on how the two options compare with each other. Faced with this situation you might well
ask yourself, “Which of these two options is the best thing for me to do now?
Which is better and which is worse?” In using the terms ‘better’ and ‘worse’ in
this way, of course, you do not mean “better from an agent-neutral impartial
perspective”. What you mean is, “Which is better as a thing for me to do at this time?” You are asking about each of
these options’ agent- and time-relative
degree of goodness, and trying to compare these agent- and time-relative degrees of goodness with
each other. If one of the two options is definitely better than the other (as a
thing for you to do at the relevant time), then it seems to be the right thing for
you to do at that time.
Broadly speaking, there are two main features that determine
each of these two options’ (agent- and time-relative) degree of goodness:
- The good feature of not diverting the trolley is that by
so doing one avoids doing anything that causes anyone to die (or to be
- The bad feature of not diverting the
trolley is that by so doing one allows a certain number of people to die.
- The good feature of diverting the trolley is that by so
doing one saves a number of people who would otherwise die.
- The bad feature of diverting the
trolley is that by so doing one does something that causes someone to die (or to be otherwise harmed).
Now it is clear that each of these good or bad features comes in degrees. I would say, for example, that one way in which the
third feature (3) can come in degrees depends on the number of people whom one will save by diverting the trolley.
Similarly, one way in which the fourth feature (4) can come in degrees depends
on the number of people whom one will cause to die or be otherwise harmed by diverting the trolley; another
way in which the fourth feature (4) can come in degrees depends on the severity of the harm that one will cause to each of these people by diverting the trolley.
As a non-consequentialist, however, I think that another way in which this fourth
feature (4) comes in degrees depends on the degree of one’s agential involvement in causing the death (or other
harm). There is, it seems to me, a whole gamut of degrees to which one may be agentially involved
in causing harm. Even though in each of these cases, diverting the trolley is
an act rather than an omission, there is still a wide
difference between different sorts of acts, depending on the degree to which
one is agentially involved in causing the harm. In general, the more deeply involved
one’s agency is in causing the harm, the worse this fourth feature (4) is.
So, at one extreme, one directly
intends the harm, and puts enormous thought
and effort into ensuring that that harm results, by carefully and
repeatedly manipulating the course of
events in order to bring about that harm. In this case, it seems to me, this fourth
feature (4) is significantly worse than at the other extreme, when one merely deflects a harmful process from one trajectory
onto another. (Intermediate between these two extremes, there are cases in which one actively sustains a harmful process,
without trying to manipulate it in any focused way. And so on.)
E.g., consider what I will call the Multiple Loop
Case (which is basically the same as a case that was first presented by Michael Otsuka in a talk at the latest Pacific APA in San Francisco): Suppose that unless the
trolley smashes into the one, it will loop round and hit the five from
other side; there is an immense spaghetti-junction of tracks, and the only way to get the trolley not to loop back towards the five is by deftly
manipulating the trolley through this complex series of junctions, in
such a way that
it hits the one person and (as a result of the impact) grinds to halt
can loop round to hit the five. I am not saying
that it is
impermissible to manipulate the trolley so that it hits the one in the Multiple Loop Case, but it does seem to
me (other things equal) a less good course of action, and so closer to being impermissible, than diverting the trolley in
the original Trolley Problem.
Another way in which the degree of one’s agential
involvement in causing a harm may vary depends on whether the primary focus of one’s intentions and
actions is on manipulating the harmful process (say, by deflecting it onto a
different path where it will do less harm), or on manipulating a person into
the path of this harmful process. When one pushes a person into the path of the
trolley, it seems to me that one is more directly and immediately involved in
causing the harm to that person than when one’s primary focus is on deflecting
the trolley onto the side track away from the five.
I don’t want to deny that this idea of "degrees of agential
involvement" in causing harm needs a lot more clarification. But I think that it
might improve our thinking about the trolley problems – and so too about
agent-relative, deontological restrictions – if we recognized that there
are many different degrees to which one may be agentially involved in causing
harm. These different degrees of agential involvement in causing harm may make
a difference to the agent- and time-relative goodness or badness of the
relevant features of the available options, and so to the overall balance of
these features that (at least in my view) determines which of these options is the right thing to