When psychological research and ethical theory are viewed from up close, it’s easy to conclude that they are utterly distinct inquiries. But this impression is mistaken.
My basic argument is simple. When we ask ourselves whether we should switch a trolley to save five, we take our intuition about this possible act to give good reasons for moral belief. We take it that
Our moral intuitions about particular cases give us defeasible reason to believe in their contents.Footnote 21
But ethical reflection doesn’t stop here. We also seek general moral principles that would explain our pattern of intuitions about particular cases. We assume that, when not biased, these intuitions systematically respond to the presence and absence of morally relevant properties. In other words,
Our moral intuitions about particular cases track certain moral principles.
Our intuitions may tell us that it’s permissible to divert the trolley in Sidetrack, but wrong to push the stranger in Footbridge. These acts, however, are surely permissible or wrong in virtue of something, and about this our intuitions are typically silent. But if there are general moral principles, and our intuitions are broadly reliable, then our intuitions should be systematically responsive to the properties cited by these principles.Footnote 22 And this means that by identifying the properties to which our intuitions systematically respond, we can identify the principles that are implicit in them. Therefore, so long as we are entitled to assume (1),
Evidence about what moral principles our intuitions track gives us defeasible reason to believe in these moral principles. [From 1, 2]
Facts about what principles our intuitions track are empirical facts, and are therefore discoverable using the methods of empirical psychology.
More specifically, facts about what our intuitions track are exactly the kind of functional facts psychologists seek to discover.
Psychological evidence about the principles our intuitions track gives us defeasible reasons to endorse these moral principles. [From 3, 4]Footnote 23
The argument is meant to apply to anyone who allows that moral intuitions about particular cases can give at least some epistemic support to general principles—to all who accept (1) and (2).Footnote 24 Versions of these premises are, I believe, accepted by many ethicists. I’ll later briefly discuss approaches that reject (1), and only extreme particularists accept (1) without accepting (2).Footnote 25 Notice that doubts about the extent to which our intuitions are reliable—doubts perhaps fuelled by empirical evidence about disagreement and diversity—only affect the scope of the argument. My argument is simply that whenever we are justified in giving weight to local armchair coherence, we should give weight to psychological evidence.
Premise (3) is the key move. It might be wondered how it squares with my earlier description of the role of coherence in ethical theorising. Indeed, it might be argued as follows: ‘Relations of coherence between intuitions and principles are logical relations, hence a priori, hence to be discovered in the armchair. If it is coherence with intuitions that supports principles, then no opening is left for empirical input to the process. It could not matter whether the principles ethicists identify converge with those identified by the psychologists. The armchair is safe.’
This objection misrepresents our interest in coherence between intuitions and principles. We accept (3) because we accept (2). That is,
We have reason to believe in the principle that best coheres with our intuitions because such local coherence is evidence that this is the principle that our intuitions track.
This understanding of our interest in coherence might seem surprising. It is therefore worth considering what it would mean to deny it.
One way to deny it is to think of our aim in ethical theorising as not to uncover pre-existing principles, but to construct general principles that would preserve as many of our particular intuitions as possible—this view sees moral principles as no more than ways of organizing or systematizing our particular intuitions, in the way scientific theories are claimed to organize experience on some instrumentalist views.Footnote 26
But such a view coheres poorly with the way most ethicists see their aims, and with actual ethical practice. When ethicists find our pattern of intuitions about trolley cases puzzling, when they wonder what principle might be implicit in it, they clearly don’t take such an instrumental attitude to moral principles. Nor do ethicists see systematic regularities in our moral intuitions as merely a fortunate accident which helps us better integrate our diverse responses to particular cases.
Most ethicists do not simply seek the principles that would preserve as many of their existing intuitions as possible. On the contrary, they actively generate intuitions about new cases, cases that might decide between competing principles that seem to cohere equally with the existing set of intuitions. This would make little sense if their aim was mere local coherence. Indeed, one possible consequence of extending the range of intuitions in the set is that we would come to reject firm intuitions from the initial set, intuitions that wouldn’t best cohere with the principle that best coheres with the larger set. Again, this would make no sense if the aim of ethical theorising was merely to preserve as many of our existing intuitions as possible. Ethicists aren’t satisfied with just any principle that happens to cohere with our existing set of intuitions—they worry that this pattern is merely accidental, that a different pattern will emerge when they extend the set.
Then there is the point that ethicists don’t admit just any intuition to the set. Intuitions must first go through epistemic screening: they must, for example, be the result of clear and calm reflection, not agitation or self-interested bias. Again it seems that ethicists aren’t just interested in preserving our existing intuitions, but in removing noise and interference that might make us perceive in them what is merely a random pattern. In ethical theorising, then, we also want to identify non-accidental patterns in non-accidental intuitions.
It might be objected that this screening is pragmatic, aiming merely at stability. We simply want to rule out transient intuitions that we will no longer have later on, or that we are less likely to share with others. But this is implausible. To see this, imagine that a mischievous demon decided to play with our intuitions. He flips a coin and on this basis makes us all respond with certain intuitions when we calmly reflect on certain examples. When these devilish intuitions are joined to our other intuitions, they accidentally best cohere with some principle—say, the DDE—which would not be otherwise favoured by our intuitions. It seems to me clear that if we discovered the random source of these common and highly stable intuitions, this intuitive support for the DDE would be defeated. Local coherence isn’t sufficient for justification when it’s known to be merely accidental.
So what we seek in ethical theorising are non-accidental patterns in non-accidental intuitions. The armchair search for local coherence is a way to detect such patterns, a way to identify the properties that our intuitions are systematically responding to, and to identify and dismiss stray or biased intuitions that obscure this underlying pattern. But this is just is to accept (3) and the understanding of the coherence that I suggested above.Footnote 27
Consider next a different objection. ‘The problem with (3) is rather that it assumes a far too simple epistemic step from intuitions to principles. When we engage in ethical theorising, we constantly move back and forth between particular intuitions and principles, as well as between these and other principles and background theories. It is this holistic character of moral theorising that your argument distorts or overlooks, and which blocks the step to (5).’Footnote 28
Premise (3) assumes that facts about the properties that our moral intuitions track give epistemic grounds for believing in the corresponding moral principles: that if our intuitions about trolley cases respond to the distinction between intending and foreseeing, then this gives us reason to endorse the DDE.
When we go back and forth between intuitions and principles, one thing we are trying to do is distinguish those intuitions that are reliable from those who are merely random or biased—and thus avoid endorsing principles based on such irrelevant intuitions. Since the aim here is precisely to single out the properties our intuitions are genuinely tracking, this feature of our practice is hardly an objection to (3). It supports it.Footnote 29
So the issue must really be about epistemic considerations that go beyond local coherence. But my argument is perfectly compatible with accepting other epistemic factors into the mix, including top-down pressures from other principles and values, and from background theories. As I emphasized, the support given by particular intuitions is defeasible. The principle they identify might turn out to be intrinsically implausible or in tension with strongly supported values. And background theories and other considerations (including empirical evidence about bias and diversity) might give us reason to think that some set of intuitions is unreliable or prejudiced, defeating premise (1) for that particular domain. None of this is in tension with my argument.
The argument is also compatible with more subtle top-down pressures. Suppose some set of intuitions track property A, but considerations of global coherence support a somewhat different property A*. Depending on the respective epistemic weight we give to local intuitions and to more global considerations, we might have overall reason to accept factor A* over A, and to revise our ground-level judgements accordingly.Footnote 30 This doesn’t show that there was no reason to adopt A, not even that the intuitional support for A didn’t play an important part in supporting A* (if more global considerations had supported some very different factor B, we might have had intuitional grounds for rejecting it). And that intuitional support for A could come from psychological evidence.
My argument, then, is entirely compatible with acknowledging top-down epistemic pressures, though I suspect that in many ethical domains, including discussion of trolley problems, they are in fact fairly weak.
The objection can therefore be defused. If it is to have real force against the argument, it must involve the claim that it’s simply
impossible to disentangle the epistemic contribution made by particular intuitions from that made by these other factors. But I doubt that this is a practicable (or even coherent) view of moral justification, and it seems to me to bear little resemblance to actual moral practice.
The step from (3) and (4) to (5) is a simple one. To reject (4) one would need to show that although our intuitions do often track general principles, that they do so is not any kind of empirical fact, indeed not something on which empirical evidence could even bear. I am not sure how one could defend such a claim, which comes dangerously close to suggesting that the empirical research actually being done in this area is incoherent.Footnote 31
The armchair search for local coherence isn’t simply a matter of assessing relations of coherence between a given set of intuitions about particular cases and our more general moral beliefs. The factors to which our intuitions are responding are typically opaque to introspection—our intuitions tell us that pushing is wrong, but not why.Footnote 32 And even in highly simplified and artificial examples such as Footbridge, there are numerous overlapping factors in play to which our intuitions might be responding. Philosophers form conjectures about these potential factors, conjectures which they then test, not only by assessing their coherence with particular intuitions and general beliefs, but also (and especially) by considering further cases. And these conjectures commit philosophers to what are in essence empirical predictions about how our intuitions would respond to the presence and absence of the conjectured factors in these further cases.
Moral philosophers certainly are happy to engage in empirical speculation about the psychological factors that underlie our intuitions once they have concluded (or strongly suspect) that our intuitions in the relevant domain are not reliable—once they have given up (1). Judith Thomson now speculates (in ways that echo Greene’s views) that the intuitive difference between Sidetrack and Footbridge is due to differences in the directness of the harm caused.Footnote 33 This is clearly a testable empirical hypothesis, a matter for psychology to determine. But this would be a testable empirical hypothesis whether or not we reject (1). What our intuitions track, and whether they track a morally relevant property, are simply different questions.