Introduction: how Parfit did not reject Act Consequentialism

In the Preface to the first volume of On What Matters, Derek Parfit states that his two masters are Sidgwick and Kant, an act consequentialist and a deontologist. Though Parfit is clearly influenced by Sidgwick, the most philosophically sophisticated of the classical utilitarians, and mentions him a few times in volume 1 (especially with reference to the dualism of practical reason), Act Consequentialism plays little role in the discussion of normative ethics in the first two volumes of On What Matters. The normative thread of these volumes is Parfit's argument for what he calls the Triple Theory. He examines three major normative theories – Kantianism, Contractualism and Consequentialism – and seeks, by examples and arguments, to show that in order to be defensible, Kantian and Contractualist theories need to be revised. He then turns to Consequentialism, focusing on the Rule Consequentialist principle: ‘Everyone ought to follow the principles whose universal acceptance would make things go best’.

Parfit argues that Rule Consequentialism, properly understood, identifies as wrong the same set of acts that are so identified by Kantianism and Contractualism. These are acts that are disallowed by principles that (1) would make things go best, (2) are uniquely universally willable, and (3) cannot reasonably be rejected. At the end of volume 1 of On What Matters, Parfit indicates that the fact that these three theories converge is important, because it is widely believed that there are deep disagreements between Kantians, Contractualists and Consequentialists. If there were such deep disagreements, adherents of each view would have grounds to doubt their own view, and even to doubt whether ‘any of us could be right’ – that is, whether there are objective truths in ethics. But, Parfit tells us, the disagreements are not deep: ‘These people are climbing the same mountain on different sides’.Footnote ¹

It would be easy to conclude from all this that Parfit believes that the most defensible form of consequentialism is Rule Consequentialism. That is in fact what we concluded on first reading volumes 1 and 2 of On What Matters. The omission of any serious discussion of Act Consequentialism disappointed us. But when Peter Singer sent Parfit a draft of his review of On What Matters for the Times Literary Supplement, we learned that our belief that Parfit thought Rule Consequentialism to be the most defensible form of consequentialism was in error. Commenting on a passage in the review, Parfit replied: ‘I don't in fact argue that the most defensible form of Consequentialism is Rule Consequentialism’. Parfit pointed to a passage in On What Matters, volume 1, in which he raised an objection to the Rule Consequentialist view that acts are wrong when they violate the rules that, if universally accepted, would make things go best. In that passage, Parfit had written:

In his response to Singer, Parfit also commented that ‘on page 418, I deliberately refrain from saying that we ought to accept the Triple Theory. I write that we ought to accept (S) and (T), but only that we have strong reasons to accept (R) and (U)’.Footnote ³

The principles Parfit refers to are as follows:

Finally, (U) is the important principle for the Triple Theory. It reads:

Summarizing what he was doing, Parfit wrote to Singer: ‘What I argue is that, if Kantianism and Contractualism are revised in ways that are clearly needed, these theories coincide with some forms of Consequentialism’.

Parfit made another comment on the draft review that is important for understanding what he is doing when he discusses Act Consequentialism in the third volume of On What Matters. Singer wrote that Parfit seemed to assume that the answer to the question ‘What ought I to do?’ is to be found by asking what principles would, if universally accepted, make things go best. Against this, Singer drew on work we had done earlier defending Sidgwick's idea of esoteric morality, pointing out that whether it would sometimes be right to torture (for example when torturing a terrorist is the only way to get him to reveal the location of a ticking nuclear time-bomb hidden in the midst of a city), and whether it is desirable for everyone to believe that torture is sometimes right, are two different questions. Singer then wrote:

To this, Parfit commented:

Parfit's discussion of Act Consequentialism in the third volume of On What Matters should be seen as developing the ‘broader form’ of convergence hinted at in this paragraph – a form of convergence between Rule Consequentialism and Act Consequentialism. We shall begin our account of it by discussing the distinction Parfit makes between Impartial Rationality and Morality.

Impartial reason as morality's rival

The conflict between egoism and morality is familiar to all of us. Sidgwick used the term ‘ethics’ in a broad sense to refer to any rational method of deciding what we have most reason to do. In this sense of ‘ethics’ egoism is a method of ethics, and the conflict between egoism and universal benevolence is a conflict within ethics. Parfit takes a narrower view of morality, and regards egoism as an external rival to morality. There is nothing new about that. But in the third volume of On What Matters, Parfit suggests that it is not only personal reasons that may be in conflict with morality. There can also be conflicts, he says, between moral and impartial reasons. ‘We should ask’, he writes, ‘whether in such cases we might sometimes have sufficient reasons or decisive impartial reasons to act wrongly’.Footnote ⁴ He answers that this view is plausible, not in its strong form, in which we always have decisive reasons to do whatever would make things go best, but in the form of what he calls:

Before we go on to consider an example, we pause to ask why Parfit raises this possibility of a conflict between moral and impartial reasons. The answer, we believe, is that it serves his goal of reducing disagreements between different normative theories. Sometimes we think we are disagreeing, but the disagreement is only in appearance, because we are giving different answers to different questions. So the egoist who offers an answer to ‘What is the rational thing for me to do?’ may not be disagreeing with the moralist who is asking ‘What is the right thing for me to do?’ But the same thing may be true of the impartial rationalist, who does not ask ‘What is the right thing for me to do?’ but rather ‘What do I have impartial reasons for doing?’ As Parfit writes, ‘Impartial Rationalism may be an external rival to our moral view, since these Rationalists may believe that, in some cases, we would have sufficient or decisive reasons to act wrongly’.

The advantage of this move, for Parfit, is that it reduces the scope of moral disagreements. If we think of Act Consequentialism as a moral theory, then it is in disagreement with any moral view that holds that it is sometimes wrong to do what makes things go best. This disagreement is internal to morality. Parfit will subsequently provide some considerations for modifying Act Consequentialism so as to reduce the extent of this disagreement. We will examine these considerations shortly. But Parfit's other strategy for reducing moral disagreement is to distinguish Impartial Rationalism from morality, in the same way as we commonly distinguish egoism from morality. Impartial Rationalists may, he says, ‘agree with [those who are not Act Consequentialists] about which acts are right or wrong’. They are answering a different question, the question of what we have impartial reasons to do. In the ticking bomb scenario, for example, an Impartial Rationalist might agree that torture is always wrong, but believe that there are sufficient, or perhaps even decisive, impartial reasons for torturing the terrorist.

If the separation of morality from impartial rationalism has this advantage for Parfit, however, it has a corresponding serious disadvantage, of which he must have been well aware. It risks reducing the significance of morality. Parfit distinguishes what he calls ‘The Moralist's Question’ ‘What ought I morally to do?’ from ‘The Rationalist's Question’ ‘What do I have most reason to do?’ and then writes:

For morality to be ‘undermined’, we assume that Parfit means that its importance as a guide to conduct would be diminished. In the case of personal or partial reasons, Parfit seems to think that morality would not be seriously undermined if we always have sufficient reason to do what morality requires, even if we did not always have decisive reason to do it. We were critical of this in The Point of View of the Universe, arguing that morality is still seriously undermined if I often have sufficient reasons to act morally and sufficient reasons to act self-interestedly, for I may then often choose to act self-interestedly, and I am not acting contrary to any decisive reasons.Footnote ⁶ Now the move to separate impartial reasons from morality undermines the position of morality even more seriously, for I may reject morality on the basis of impartial reasons as well as on the basis of personal and partial reasons, and in doing so, I am still not acting contrary to decisive reason.

In addition, we can doubt whether, if Parfit were to succeed in distinguishing a sphere of impartial reasons, separate from morality, this would help his overall project of reducing the scope of moral disagreement. Many moral theorists take the fact that morality provides overriding reasons for action as an important or even essential element of the concept of morality. Unless Parfit can convince these theorists that they are wrong, his separation of morality from impartial reasons seems doomed to – at best – close one area of disagreement while opening up another.Footnote ⁷

Another troubling question can also be asked: when morality is opposed by impartial reason, why do we even have sufficient reason to act morally? Parfit offers this example of a conflict between morality and impartial reason:

The Nuclear Policy involves terror bombing, deliberately killing civilians. Their deaths are aimed at, as a means of persuading the Japanese government to surrender. In the Conventional Policy, the deaths of the civilians are foreseen but unintended side-effects. The president therefore has a distinctive deontological reason to choose the Conventional Policy, but he has an impartial reason to cause 200,000 fewer people to be killed. Parfit acknowledges the difficulty of comparing such different reasons, but adds:

We agree that this conclusion should be disturbing. To acknowledge that, on an issue as important as this, one can have a decisive impartial reason to act in a way that is very wrong must diminish the importance of morality. We would argue that what this move gains in making it easier to reconcile conflicting moral views is lost in making morality less important as a guide to conduct.

In a section called ‘Wrongness and Reasons’, Parfit indicates how he would have responded to this point. He repeats his earlier statement that the question about reasons is more fundamental than the question about morality, but adds that we have good reasons to go on asking the Moralist's Question, saying:

He goes on to say that we can ask, first, whether there are ‘morally decisive reasons not to act in some way’. If there are, we can then ask the further question ‘whether these morally decisive reasons are decisive all things considered, by outweighing any conflicting non-moral reasons’.

That may be true, but it invites the question whether these distinctive beliefs about which acts are wrong and what makes them wrong, and the morally decisive reasons to which they give rise, are beliefs we ought to continue to hold, if they do not give us decisive reasons to act.

Parfit concludes the section by writing:

This remark leads to our next section.

Deontic badness

We now turn to Parfit's discussion of Act Consequentialism as a moral theory, that is, as an internal rival to all moral views that hold that, in some cases, it would be wrong to do what would make things go best. This discussion is in chapters 55–58 of On What Matters, volume 3.

Parfit's overarching aim is to show that we are able to ‘resolve at least some of the disagreements between Act Consequentialism and the pluralist moral views that most of us accept’.Footnote ¹² This would, of course, bring Act Consequentialism closer to our ordinary conceptions of morality, and therefore, to the Triple Theory. One way in which Parfit seeks to advance this aim is through the distinction between deontic and non-deontic badness: ‘Acts are deontically bad’, he suggests Act Consequentialists might say, ‘when and because these acts are wrong’. All other kinds of badness are non-deontic.Footnote ¹³

This view requires Act Consequentialists to hold that it is intrinsically – and non-deontically – bad to treat people in certain ways, such as deceiving or coercing them, and when these acts are wrong, because the intrinsic badness of the deceit or coercion is not justified by other good effects of the act, ‘their wrongness makes them also deontically bad’.Footnote ¹⁴

We accept that this is a possible position for a pluralist consequentialist to hold. But how plausible is it? Parfit offers two reasons for thinking it plausible. The first, which he claims even a utilitarian could accept, involves denying that well-being needs to be limited to conscious states like happiness or pleasure, or to the satisfaction of desires. It can, Parfit suggests, include our achievements, and also our moral goodness or lack thereof. The second invites us to imagine two different worlds in which everyone has the same overall level of well-being, but everyone in the first world is morally good, and everyone in the second world is morally bad. If we accept that moral badness makes someone's life go worse, we can specify that, independently of that, the well-being of those in the second world is sufficiently better in non-moral ways to compensate for this. Parfit, following Ross, suggests that we might believe that the existence of the morally bad people would be less good than the existence of the good people. We might even think this to be the case even if the morally bad people had a higher level of well-being.Footnote ¹⁵

These reasons do not persuade us that consequentialists should include deontic badness or goodness as intrinsic values. We hold that well-being is limited to conscious states like happiness or pleasure, or to the satisfaction of desires. In fact, we agree with Sidgwick that it is ‘Desirable Consciousness which we must regard as ultimate [or as we prefer to say, intrinsic] Good’.Footnote ¹⁶ Deontic norms are instrumentally valuable when they conduce to this intrinsic good, and otherwise not.

On Parfit's second reason for holding that moral goodness and badness are intrinsic values, we share the intuition on which his example relies. That is, if we are asked to say whether we feel more positively about the world in which everyone is morally good than we do about the world in which everyone is morally bad, we acknowledge that this is the way we feel, even though the level of well-being is the same. But we regard this intuitive response as a tenuous thread on which to hang a judgement about what is of intrinsic value. Our moral intuitions are shaped by, among other things, the way in which we are taught what is right and what is wrong. As we will see in the next section, consequentialists have long acknowledged that the consequences of teaching people to follow simple moral rules may be better than the consequences of teaching them to calculate, with each choice they face, which act will have the better consequences. One way to teach and motivate people to follow moral rules is to inculcate in everyone, from an early age, the belief that certain acts are right and people who do them deserve to be rewarded, whereas other acts are wrong and people who do them deserve to be punished. That teaching, we suggest, is why most of us have the intuition that a world with morally good people is intrinsically better than a world with morally bad people in it, and why raising the well-being of those who are morally bad beyond the level of those who are morally good does nothing to assuage the feeling that the world with morally good people is a better world. (At least, this is likely to be our intuition if we assume that well-being in these worlds is positive.) Note, though, that this is an explanation of our intuitions, not a justification of them. The explanation debunks the intuition that there is intrinsic goodness or badness in certain acts, irrespective of the consequences of those acts.Footnote ¹⁷

Act Consequentialism as self-effacing

In the final chapter of the third volume of On What Matters, Parfit returns to the possibility of a normative theory being self-effacing, which he first discussed in Part I of Reasons and Persons. He points out that many Act Consequentialists assume that things would go best if we always tried to do what could be expected to make things best, but this assumption may be false:

Even just being motivated along the lines that pure Act Consequentialism implies might cut us off from some of the greatest sources of happiness, for ourselves and others:

Parfit admits that this is not universally true: if rich people, who are hundreds or thousands of times wealthier than the poorest people in the world, were to follow Act Consequentialism, and give away most of their wealth, that would make things go best. But for many other people, he thinks, acceptance of Act Consequentialism would, on the whole, make things worse.Footnote ¹⁹

These ideas are not new. Sidgwick thought that:

This was, he thought, a ‘necessary evil’ given the ‘seductive impulses’ that lead us away from morality, and the varying degrees of moral and intellectual development to be found in the world.

Parfit quotes most of the above passage from Sidgwick, and notes that Bernard Williams objected to Sidgwick's view on the grounds that it is elitist and paternalist. He responds that this objection might be justified if, as Williams believed, there are no moral truths. On that point, Parfit's claim can be questioned, given that Williams’ objection seems to be a moral one. If there are no moral truths, what significance is there in objecting to a view on the grounds that it is elitist and paternalist?Footnote ²¹ On the other hand, if there are moral truths, as Sidgwick believed, then we agree with Parfit when he goes on to say:

Parfit immediately follows this, however, with the suggestion that if Act Consequentialism is self-effacing in this way, that counts against it:

As this sentence occurs at the end of a section, it isn't immediately clear in what way Parfit thinks that Act Consequentialism's partial self-effacement might help to show that some other moral view is true. In the following section [sec 183], Parfit introduces Rule and Motive Consequentialism:

Parfit argues that, in part for reasons already mentioned, the optimific motive is not that of always wanting and trying to act in optimific ways. Instead, ‘things would go on the whole better if, rather than always wanting and trying to act in optimific ways, we had certain other motives and tried to follow certain other rules’. Thus Act Consequentialism and Rule and Motive Consequentialism are incompatible. If things would on the whole go best if most of us had these other motives – that is, motives other than the motive of trying to do the most good – and tried to follow rules other than the rule of doing what can be expected to bring about the most good, then Rule and Motive Consequentialism would be the optimific moral view. Parfit then concludes this section with the comment: ‘This fact would give us some reasons to accept Rule and Motive Consequentialism, but it would not show that Rule and Motive Consequentialism is true. We should not assume that an optimific view must be true’.Footnote ²⁴

An initial point to make here is that Parfit writes of reasons to accept Rule and Motive Consequentialism, but accepting a theory is not the same as adhering to it, and it is possible that we could accept Rule and Motive Consequentialism, in the sense of believing it to be the true moral theory, without having optimific motives and without trying to follow optimific moral rules. So from the fact that if we did have optimific motives we would do more good than if we were Act Consequentialists, it does not follow that we have some reason to accept Rule and Motive Consequentialism. This would only follow if the link could be made between believing the theory to be true, and adhering to it. Similarly, it does not follow from the fact that we accept that we ought to try to follow optimific rules that we will actually follow them. So the reasons that Parfit has offered for accepting Rule and Motive Consequentialism are no stronger than the reasons for believing that these links hold.

If we focus not on what moral view would have optimific consequence if everyone accepted it in the sense of believing it to be true, but rather on what moral view would have optimific consequences if everyone succeeded in adhering to it, then it is not clear that that view would be Rule and Motive Consequentialism. It too might be partially self-effacing. Perhaps adherence to Common Sense Morality would have better consequences, overall, than adherence to Rule and Motive Consequentialism, because if everyone accepted the latter, many of them might persuade themselves that convenient violations of the common sense rules are justifiable, even when those violations are not in fact optimific. If Rule and Motive Consequentialism may also be self-effacing, however, then the fact that Act Consequentialism may be self-effacing cannot give us any reasons to accept Rule and Motive Consequentialism rather than Act Consequentialism.Footnote ²⁵

Moreover, given the distinction Parfit himself makes between a moral theory being optimific and it being true, we are still lacking an explanation of the final sentence of the preceding section, where Parfit said that the fact that it might be better if most people rejected Act Consequentialism ‘might indirectly help to show that some other moral view is true’.

Parfit begins the next section – section 184 – by referring to the passage in the first volume of On What Matters that he quoted in his comments on Peter Singer's draft review, and which we have already cited, but will cite again here:

Parfit now says: ‘These claims made a bad mistake. What I called this “strong objection” applies only to some versions of Rule Consequentialism’.Footnote ²⁶ Parfit then refers to other familiar reasons why one might follow Rule Consequentialism rather than try to calculate what would make things go best – for example, detectives in some counter-terrorist police force might have good reasons to follow the rule ‘Never use torture’ because they are too ready to believe that the benefits of using torture outweigh the costs. This is still compatible with Sidgwick's defence of esoteric morality, for example with the idea that it would be best if most people accept some improved version of Common Sense Morality, while a few accept Act Consequentialism. But Parfit now introduces a new assumption:

This enables him to ask whether things would on the whole go better if most of us were pure Act Consequentialists, or we accepted and tried to follow certain other moral rules. From here, a path opens towards the latter, Rule Consequentialist, choice. We will not follow that path in detail. Some of it, for example the valuing of deontic goodness and badness, we have already discussed. Other aspects, such as the ability of Rule Consequentialism to overcome the problem of many imperceptibly small effects leading to great harms, are worthy of more detailed discussion than is appropriate here. Instead, we want to question the assumption that everyone ought to have the same moral beliefs. Parfit supports this assumption by saying that moral truths are not true only for certain people. Indeed, no truths are ‘true only for certain people’. This common phrase is just a misleading way of putting certain facts. I may say, for example, that it is true for me (Katarzyna de Lazari-Radek, a philosopher working at the University of Lodz in 2020) that strawberries are more desirable than raspberries, and true for you (Peter Singer, a philosopher working at Princeton University in 2020) that raspberries are more desirable than strawberries. But it is true for everyone that the specific Katarzyna identified above likes strawberries better than raspberries and true for everyone that the specific Peter identified above likes raspberries better than strawberries. Still, what does this have to do with whether everyone ought to have the same moral beliefs? Parfit has just been arguing that if most people believe that Rule and Motive Consequentialism, or perhaps some form of improved Common Sense Morality (which might turn out to be identical with Rule and Motive Consequentialism), is the true moral view, this belief is likely to have better consequences than if most people believe that Act Consequentialism is the true moral view. But we take it that the ought in ‘Everyone ought to have the same moral beliefs’ is a practical ‘ought’, and a consequentialist one at that. It is not an epistemic ‘ought’, and it is not supported by the claim that ‘Moral truths are not true only for certain people’. So we think that the path that opens for Parfit when he makes the assumption that everyone ought to have the same moral beliefs is a path that leads him away from a discussion of which is the true moral theory, and into a discussion of which moral theory it would be optimific for everyone to accept. If his claim is that this is an improved form of Common Sense Morality, and that this coincides with some form of Rule Consequentialism, we are not concerned to oppose this claim. But this does not ‘help to show that some other moral view is true’. Against that claim, we need only to repeat what Parfit has himself said: ‘We should not assume that an optimific view must be true’.Footnote ²⁸