Act and Rule Consequentialism: A Synthesis

4.1 Two Accounts of Value and Reasons

The first axiology we need for the two new hybrid theories is the traditional agent-neutral theory of value, which the rule consequentialists have traditionally used for evaluatively ranking the outcomes of the general adoption of different moral codes.^[13] According to rule utilitarianism, how good an outcome is depends only on the total amount of wellbeing it contains. More recently, rule consequentialists have insisted that the goodness of an outcome may also depend more on how well off the worst-off individuals are and/or the amount of other goods such as fairness, biodiversity, and/or knowledge. Here the selection of the most plausible axiology depends on how well the resulting versions of rule consequentialism can fit our moral convictions and how unified and explanatory the resulting account will be (Hooker 2000, ch. 1 and Sections 2.3–2.5). For our purposes, it does not matter exactly which prima facie plausible agent-neutral axiology we adopt at this point. The axiology we need is thus a previous type of an ordinary rule-consequentialist account of agent-neutral value, call it value_a-n.

The second axiology we need must be agent-relative. Such axiologies are common in the so-called ‘consequentialising’ literature, where they are used to formulate versions of act consequentialism that can accommodate agent-centred constraints.^[14] These versions of act consequentialism agree that we are not always permitted to do whatever makes things go best agent-neutrally, because our own agential involvement (such as our actions of harming or killing others) can sometimes make the resulting outcomes worse relative to us but not relative to others. What is distinctive about such agent-relative axiologies is that the very same aspect of an outcome can be good relative to one agent but either neutral or bad relative to others.

The agent-relative axiology we need has two essential features. Firstly, it needs to accept that whatever qualities (such as general wellbeing) make an outcome agent-neutrally good according to the previous axiology (that is, whatever has value_a-n), those very same features also make outcomes good relative to every agent. Secondly, it furthermore needs to recognise that there are also other considerations related to the agent, her agential involvement, and especially her rule-following and character traits that can also directly affect how good an outcome is relative to her as an individual.

What are these considerations? To answer this question, we can begin from the idea that sometimes when you act, complying with a specific moral principle can be the end of your action.^[15] Hence, the end of my action of handing over a book to my friend can, for example, be to comply with the principle that borrowed books must be returned. Furthermore, as Christine Korsgaard (2005, 22) has argued, the end of an action can be thought to constitute in part the identity of that action – it can make the action the action it is. Giving a book to a friend as a birthday present and giving a book to her to comply with the previous moral principle are, after all, intuitively different actions just because they have different ends.

What I then want to suggest is that it can be good relative to an agent that she has done an action that has as its constitutive end the aim of complying with a moral principle that belongs to the optimific code. That the agent has done such an action would, on this view, be something that can itself make an outcome better relative to her. If what is right were then a function of what is the best relative to you, the previous axiology would often entail that you should follow the optimific code even in the divergence cases.

However, to make the previous simple suggestion compelling, a substantial explanation must be provided of just why it would be good relative to an agent that she has done an action that has as its constitutive end a moral principle that belongs to the optimific code. I believe that such an explanation can be provided if we begin from Brad Hooker’s (2002) own rule-consequentialist account of moral virtue instead of from an account of the independent moral value of certain character traits.

According to Hooker (2002), there are two necessary requirements that are jointly sufficient for a given character trait to count as a virtue. By character traits, Hooker (2002, 23) means settled dispositions to act, think, and react in certain ways. These dispositions thus contain behavioural elements, but also motivations, seeing things in a certain light, emotions and reactive attitudes, and ways of reacting to different actions with those attitudes.

Hooker’s (2002, 38–9) first condition for whether some set of the previous types of dispositions (i.e. a character trait) counts as a virtue is whether the agent having that character trait would be good for everyone. That is, the test is whether having those dispositions that constitute the trait have good consequences for wellbeing generally. Following Hursthouse (1999, 204), Hooker (2002, 34) more precisely emphasises that the trait must have the good consequences in ‘a favourable environment’, as having virtuous dispositions must be ‘standardly’ beneficial for ‘normal’ human beings. Even more importantly, Hooker (2000) also stipulates that ‘what makes an environment a favourable one is that the other people in it are virtuous and have strongly negative reactions to those who aren’t (34).’

We can then make two observations about this necessary condition (hereafter the optimality test). Firstly, Hooker (2002, 28 fn. 14; 2000, 91) understands the adoption of a moral code in terms of developing exactly the previous kinds of dispositions to behave, think, and react in certain ways (that is, in terms of developing a moral conscience with a certain shape). This means that (i) adopting the ideal code in the rule-consequentialist sense and (ii) possessing the character traits that meet the previous optimality test for the status of being a virtue are for him one and the same thing. In other words, the content of the optimific code and the content of the dispositions that meet the first test is identical. This is why Hooker (2002, ch. 3) argues that the most plausible form of virtue ethics in the end collapses into rule consequentialism.

All of this entails crucially that the connection between the optimific code and character traits that count as virtues will not be merely accidental on this view, but rather the optimific code provides virtues the content they must have in order to count as virtuous character traits in the first place. This internal connection makes the optimific code an indispensable element of the hybrid view described below. And, even more importantly, the status of being a virtue is on this view ultimately and fundamentally tied to what is of value according to the previous agent-neutral axiology, namely wellbeing, and so in this respect the proposal described in this section will not lead to introducing a third axiology that would be independent of the other two axiologies introduced in this section.

Yet, according to Hooker, it is not sufficient for being a virtue that a given character trait passes the optimality test. Rather, the trait must also be actually desirable for both the agent herself and also for other people around her. This is because being a ‘bad virtue’ would be a contradiction in terms (Hooker 2002, 23). The character traits that meet this desirability test can be desirable already because of their instrumental value. Trustworthiness, for example, is a trait most happy people tend to have, as the relationships it enables are both an important source of pleasure and constituents of human wellbeing themselves. Furthermore, the traits that meet the optimality test are already desirable for others too because they can feel safe around those who possess them, they can trust those people, and like them more easily too. These traits are thus arguably desirable because they create an atmosphere of trust and mutually beneficial co-operation that leads to general wellbeing (Hooker 2000, Section 4.2).^[16] They can also be said to be desirable, for example, because developing them is an achievement which itself is an important personal good (that is, a constituent of an individual’s own wellbeing) (Hooker 2002, 24).^[17]

I will assume that a character trait counts as a virtue if and only if it passes the optimality and the desirability tests. We can then finally formulate the required agent-relative axiology. Thus, according to the resulting agent-relative value theory, (i) any consideration that makes an outcome agent-neutrally good also makes it agent-relatively good and (ii) in addition it is good relative to an agent that she does an action that exemplifies a virtuous character trait that is both optimal in the favourable circumstances and desirable for the agent herself and others in the actual circumstances. Let us call this value value_a-r.^[18]

Value_a-n and value_a-r can then be used to give two corresponding accounts of practical reasons. As consequentialist accounts, these accounts take practical reasons to be determined by the value of outcomes. The suggestion is that we should take the previous two axiologies to provide accounts of the reasons which different subjects have. We first use value_a-r to give an account of the practical reasons of ordinary situated agents. On this view, we in the real world have reasons to bring about an outcome to the degree that the outcome contains value_a-r, that is, insofar as the action brings about (i) things that make the outcome agent-neutrally good and (ii) doings of actions which exemplify virtuous character traits.^[19]

Value_a-n can, by contrast, be used to give an account of the practical reasons of an impartial observer. An impartial observer is a theoretical construct of a person who has no personal ties (and so no partialities) and who can never do anything (and so her own virtuous agential involvement cannot affect the value of outcomes).^[20] We can then think that how strong reasons an impartial observer has for preferring an outcome is determined by how much value_a-n the outcome contains. As a consequence, how much agent-neutral value the outcome of the general adoption of a moral code contains determines how strong the reasons are for the impartial spectator to prefer the general adoption of that code (see Parfit 2011, 372).

We can then make two observations. Firstly, the previous two axiologies are fully consistent with one another. It is true that value_a-r recognises one source of value that value_a-n does not, but this can be explained, because the function of these two value properties is to provide an account of the reasons which different subjects have. The function of value_a-r is to give an account of the reasons of ordinary situated agents, and so this value can be sensitive to how good, relative to these agents, their actions are in exemplifying virtuous character traits. In contrast, because the purpose of value_a-n is to give an account of the impartial observer’s reasons it simply cannot take such an agent-relative value into account, as the impartial observer never does anything.^[21]

Secondly, because these two axiologies and the reasons they ground are consistent with one another, the hybrid view I will formulate on the basis of them will not be vulnerable to the incoherence objection. Furthermore, value_a-r also provides a substantial, virtue-based account of what can make it good, relative to an agent, that she does an action that has an optimific principle as its end. This account is based on the two tests, closely connected to wellbeing, for why certain character traits such as honesty, beneficence, and so on count as virtues. For this reason, the view introduced below will never ask us to comply with the optimific moral code just for the sake of it, and so the view can avoid the rule-worship objection too.

4.2 A New Hybrid Version of Consequentialism

We now have the resources required for a new hybrid version of consequentialism that combines elements of both act and rule consequentialism. The key innovation of this view is that it gives the optimific code a different role. In this theory, that code is not used to explain how rightness relates to good outcomes but rather to explain value, virtue, and the agent-relative reasons associated with virtuous actions.

We can call the new view hybrid consequentialism. This view rejects the idea that right and wrong are a function of what is authorised by the optimific principles, and so in this sense it is not a rule-consequentialist view. Yet, the first positive element of hybrid consequentialism is an account of practical reasons, which combines elements of both act and rule consequentialism. It is act consequentialist in the sense that what an agent has most reason to do (and what she ought to do) is determined directly by which action brings about most (agent-relative) value (i.e. value_a-r). It is, however, also rule consequentialist in the sense that the optimific rules can affect how good the outcome of a given action is relative to the agent. This is because, as explained, that an action with a certain optimific principle as its end has been done can be good relative to the agent insofar as doing the action thereby exhibits a virtuous character trait that passes both the optimality and desirability tests. On this view therefore, the optimific code can affect how good the consequences of different actions are relative to the agent, and so also in part determine what the agent has most reason to do and what she ought to do. A rule-consequentialist element is thus also built into this view’s act-consequentialist theories of value, reasons, and ought. On this view thus both what reasons an agent has and what she ought to do is determined by how much value_a-r her actions bring about.

The second element of hybrid consequentialism then is that, on this view, which actions are right and wrong is also determined by what maximizes value_a-r. Thus, according to hybrid consequentialism, an action is right if and only if it brings about at least as much value_a-r as any other action available for the agent. Yet, even if this view is act-consequentialist in structure in this way too, it is not a version of pure act consequentialism, because according to it an agent’s reasons (what she ought to do) and which actions are right and wrong all still partly depend on the optimific principles. This is because on this view that an agent does an action that has an optimific principle as its end can itself make the outcome of that action good relative to the agent, and so such principles can affect reasons, what one ought to do, and what is right and wrong.^[22] Here the rule-consequentialist element of the view is again built into the view’s theories of value, reasons, and ought, and thereby even into the view’s account of right and wrong actions.

Let me conclude this section by making one observation about hybrid consequentialism. One initially attractive feature of the view is that given how the view is formulated, hybrid consequentialism entails the strong moral rationalism thesis. Strong moral rationalism is, roughly, the view that we always have decisive reasons to do what is right and to avoid actions that are wrong.^[23] Hybrid consequentialism endorses this claim because, according to it, an action that is right must always be an action that you have most reason to do. This is because both rightness and what you have most reason to do is, on this view, determined by what brings about most value_a-r.

We have then a new type of a hybrid-consequentialist theory on the table. The final question we thus need to consider is whether it can provide a response to the ideal-world objection.

4.3 The Hybrid Theory, the Ideal-World Objection, and the Divergence Cases

This section returns to the divergence cases of Section 2 by investigating whether hybrid consequentialism can match our convictions about these cases.

There seems to be two kinds of divergence cases. What is distinctive about Boat, Fishing, and Gremlin is that we intuitively should not comply with the optimific code in them. In Boat we should throw additional lifejackets, in Fishing we should use a bigger net, and in Gremlin not do a jumping jack at noon even if these actions are forbidden by the optimific code. By contrast, even if Transplant and Promise are divergence cases, we intuitively ought to follow the optimific code in them. Thus, in Transplant we ought to not operate and in Promise we should take our own children to the park even if there are actions that would have agent-neutrally better consequences.

Let us consider the first kind of cases first, starting from Boat. To see what implications the hybrid view has in Boat, we first need to compare how much value_a-r the outcomes of different actions would contain in it. In this case, you can either throw just one lifejacket or continue throwing more lifejackets to save more lives. If you throw just one lifejacket, the consequences are that (i) one life is saved, (ii) nine people die, and (iii) you have done an action the end of which is to comply with a life-saving principle that belongs to the optimific code. It can then be argued that doing an action that has this principle as its end is good relative to the agent because it exhibits the trait of doing your fair, equal share, which perhaps could be suggested to be a virtuous trait. This is because this trait seems to pass both (i) the optimality test and also (ii) the desirability test (because having the trait of doing your fair share is, for example, conducive to an atmosphere of smooth, predictable co-operation). Yet, the consequence of throwing additional lifejackets is that all 10 people are saved, and so arguably here throwing additional lifejackets brings about more value_a-r than following the optimific code even when we consider the value of acting from virtuous character traits.

This means that in Boat hybrid consequentialism entails that you have most reason to keep throwing additional lifejackets, that you ought to do so, and that this action would also be the right thing to do. Hybrid consequentialism can thus match our moral convictions about this case insofar as throwing additional lifejackets intuitively is the right thing to do.

Hybrid consequentialism has similar consequences in Fishing and Gremlin. In Fishing, the view can claim that even if using small nets is good relative to the agent because it has an optimific principle as its end and so the action exhibits a virtuous character trait (of, say, caring about the environment), the outcome of using larger nets still has more value_a-r given all the agent-neutrally good consequences that follow from doing so. It follows that using larger nets is, according to hybrid consequentialism, what you have most reason to do, what you ought to do, and the right thing to do, which again is the intuitively right verdict.

In Gremlin, having the relevant optimific moral principle that requires doing a jumping jack at noon as an end of your actions cannot, however, be claimed to constitute a disposition that would count as a virtuous character trait. It can, of course, be argued that having this character trait promotes general wellbeing in the favourable circumstances in which others share it, and so the trait passes the optimality test. However, having this trait is not desirable for the agent or for others in the actual circumstances in which others do not share it (it merely produces headaches for everyone) and so the trait fails the desirability test. This means that doing a jumping jack at noon with the optimific jumping-jack principle as its end cannot make the outcome of that action any better relative to the agent in the real world, given that the agent would not be acting from a virtuous character trait. As a consequence, hybrid consequentialism entails that you have no reason to do a jumping jack at noon when others have not internalised the jumping-jack principle, and so it is not what you ought to do or the right thing to do.^[24]

What about the divergence cases in which we intuitively ought to follow the optimific principles? In Transplant, it can be argued that when you refuse to operate, you have as an end of your action an optimific ‘Do not kill!’ principle. It can also be argued that having such a principle as an end constitutes a virtuous character trait – the trait of treating each person as an end in themselves, which could be said to fall under the personal virtue of justice.^[25] This trait passes the optimality test because if most people lacked it in the society we would be too anxious to see a doctor, which would have catastrophic consequences (Parfit 2011, 363). Having this trait also passes the desirability test because in the real world it enables us to feel safe in knowing that we will not be sacrificed for the sake of others. This trait thus seems to pass Hooker’s two tests for virtuous character traits.

To reach the intuitive verdict that we ought not to operate in Transplant, hybrid consequentialism would then have to insist that not killing the orphan, merely letting the five individuals die, and doing an action that exhibits the virtuous character trait is better relative to the agent than killing the orphan and saving the five. With this axiology, the defenders of hybrid consequentialism can argue that not operating maximises value_a-r and so it is what you have most reason to do, what you ought to do, and the right thing to do.

Finally in Promise, hybrid consequentialism can recognise that acting from the optimific promise-keeping principle exhibits the virtuous character trait of trustworthiness and so the outcome of keeping your promise and taking your own children to the park is better relative to you than taking your neighbour’s children. Hybrid consequentialists can thus claim that keeping the promise is what you have most reason to do, what you ought to do, and also the right thing to do.^[26]

Overall, this means that hybrid consequentialism seems able to match our moral convictions about the divergence cases we examined. The point of these cases was originally to illustrate the ideal-world objection by showing that rule consequentialism is extensionally inadequate – that due to structural reasons rule consequentialism has implausible ethical consequences. Given that hybrid consequentialism matches our convictions in these cases, the view thus seems able to avoid the ideal-world objection too.