Abstract
Like most philosophical terms, ‘consequentialism’ has no universally accepted definition. In this chapter, I shall try to arrive at a characterization of the doctrine of consequentialism that fits present-day usage reasonably well. I shall also argue, rather briefly, for the normative plausibility of some of the theses included in this characterization. Partly, this argumentation will take the form of criticism of alternative conceptions of consequentialism. Thus, I shall for example discuss Michael Slote’s ‘satisficing’ consequentialism (in 2.4), as well as a form of ‘probabilistic’ consequentialism, defended by Frank Jackson (in 2.5).
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
I shall treat the phrases ‘is obligatory’ and ‘ought to be done’, concerning actions, as synonymous. According to some moral theories, there are further kinds of moral status, for example supererogatoriness, besides obligatoriness, Tightness, and wrongness. Consequentialism is often thought to entail that there are no such further kinds of moral status. (Cf. 2.4.)
Two problems for this approach are how to individuate generic actions, and how to distinguish them from other properties. A discussion of these problems can be found in Goldman(1), pp. 12–19.
This is a rather common account of particular actions. See e.g. Von Wright, pp. 36–37, Bergstrom(l), p. 21, and Goldman(l), p. 10. Goldman observes that a fourth determinant, besides agent, generic action and time(-interval), may be necessary to make this criterion of individuation satisfactory. (Goldman(l), p. 11, note 13.) A person may for example make a certain signal twice at the same time, once with his left hand and once with his right. We would then be inclined to say that he has performed two particular actions. (Goldman suggests adding the way in which an action is performed, as a further determinant.) I shall ignore this complication, however.
Cf. Bergström(I), p. 24, and Goldman(I), p. 47.
Cf. Bergström’s discussion of compound actions and ‘activities’ in Bergström(I), p. 26.
Cf. Bergström(I), pp. 32–33.
Some philosophers claim that all actions are intentional, as a matter of conceptual truth. (See e.g. Von Wright, p. 36.) I shall assume, however, that there are unintentional actions.
It is explicitly assumed e.g. by Samuel Scheffler in Scheffler(1), p. 1, and in Scheffler(2), p. 1.
The quasi-objective conception of intrinsic value may seem incompatible with both the ‘relational’ and the ‘subjective’ conception of intrinsic value, as we may call them. According to the relational conception, a state of affairs cannot have intrinsic value simpliciter, it can only have intrinsic value for somebody. (Cf. Oddie’s and Milne’s distinction between ‘agent neutral’ and ‘agent relative’ value (Oddie & Mllne, p. 42), as well as Jan österberg’s distinction between the ‘objective’ and the ‘subjective’ conception of intrinsic value (öSTERBERG, p. 44). What österberg calls ‘the subjective conception’ is what I call ‘the relational conception’.) According to the subjective conception, which is compatible with the relational conception, a state of affairs cannot have intrinsic value simpliciter, it can only have intrinsic value according to somebody. Even if intrinsic value is ultimately relational or subjective, however, a ‘non-relational’ and ‘objective’ concept could be defined, by stipulating that a state of affairs has intrinsic value or disvalue simpliciter iff it has intrinsic value or disvalue for somebody, or according to somebody. (If this non-relational and objective concept is to be of any use, we must presuppose at least some degree of interpersonal comparability. A state of affairs may have different intrinsic value or disvalue for, or according to, different persons. Also, a special problem arises if the ‘resultant value’ of a certain state of affairs (got by aggregating its values or disvalues for or according to different people) is ‘zero’ or ‘neutral’. If we deny that such a state of affairs has any intrinsic value or disvalue simpliciter, the ‘if-clause of the proposed definition is violated. A possible solution would be to distinguish between states of affairs that lack intrinsic value or disvalue simpliciter, and states of affairs that have neutral intrinsic value simpliciter.)
By the ‘relative’ intrinsic value of a′s outcome I mean its value as compared to the values of the outcomes of alternatives to a. The problematic notion of alternative actions will be discussed in Chapter 6. For the moment I shall just suppose that it is always possible, at least ‘in principle’, to determine whether two particular actions are alternatives to each other.
The label ‘axiological theory’ is taken from Vallentyne(I), p. 25.
There are some theories, naturally regarded as consequentialistic, which are neither factualistic nor probabilistic. One such example is Fred Feldman’s theory, discussed in 7.5.
A few more remarks about the philosophical significance of AX are made in 3.2.
See e.g. Scheffler(1), p. 1, Kagan, p. xi, and Sosa, p. 101.
Given the quasi-objective conception of intrinsic value, ethical egoism, e.g., cannot be formulated as a totalistic theory. This conception of value precludes the egoist from claiming that the fact that a certain (total) outcome is better for the agent than another outcome is a sufficient condition for its being better on the whole. (The egoist can nevertheless be an axiologist, and accept the quasi-objective conception of intrinsic value, by rejecting totalism and holding that the morally relevant outcome of an action only contains states of affairs which concern the agent herself. Cf. Chapter 3.)
Provided that CEM is true. See Below.
See e.g. Oddie & Milne, p. 53.
Moore(l), pp. 27–31. See also Chapter 4, below.
Cf. Bergström(I), pp. 64–65. This definition of ‘future tate’ may seem to rule out the idea, sometimes said to be characteristic of consequentialism, that an action’s own intrinsic value is never relevant for its moral status. As Bergström remarks, however, this idea could be reintroduced by means of some ‘extra requirement’ on outcomes. (Bergström(l), p. 65.)
Moore(3), p. 559. Moore seems at other times to have interpreted outcomes in accordance with WO or FO.
Bergström(1), Chapters 3 and 4. Bergström is concerned with finding an interpretation of ‘consequences’ that, besides being normatively acceptable, is ‘descriptively reasonable’ in the sense that it does not deviate too much from the ordinary usage of the term. Since ‘outcome’ does not seem to have the same causal connotations as ‘consequences’, he would perhaps not insist that the interpretation of the former term must include any reference to the concept of causality. (It may be thought misleading to use the label ‘consequentialism’ for theories that do not deal with ‘consequences’ in any ordinary sense. But such is the current practice, and ‘outcomism’ does not sound very good, as David Sosa observes. (SOSA, p. 101.))
I shall assume a possible-worlds analysis of counterfactual conditionals, in the style of Robert Stalnaker (Stalnaker(l)) and David Lewis (Lewis(1), Lewis(3)). According to Stalnaker’s theory, which is the simpler one, the counterfactual A>B is true (in the actual world) iff B is true in the nearest or most similar possible world where A is true. In Stalnaker(l), Stalnaker assumes that, for every proposition A, there is a unique nearest possible world where A is true. This is, on his theory, equivalent to assuming CEM. (Due to criticism from Lewis (Lewis(1), pp. 79–80), Stalnaker has later abandoned CEM.) Henceforth I shall use the term ‘counterfactual’ as short for ‘counterfactual conditional’, and I shall often use ‘world’ as short for ‘possible world’. Despite its name, a counterfactual need not have a false antecedent. On Stalnaker’s and Lewis’s approach, a counterfactual with a true antecedent is true iff its consequent is true. (See Lewis(I), p. 26.) I will say nothing about how the relation of nearness or similarity between worlds is to be understood (see LEWIS(I), pp. 91–95 and Lewis(3)), and I do not wish to commit myself to any particular view about the ontological status of possible worlds. (Lewis regards possible worlds as concrete entities of the same ontological type as the actual world (Lewis(I), pp. 84–91, Lewis(4)), while Stalnaker is a more moderate realist. (Stalnaker(2).))
If Cem is rejected, the ‘principle of Subjunctive Outcomes’: SO The outcome of an action is the states of affairs which would occur if the action were performed, would constitute an interesting fourth view of outcomes. A fifth variant could be defined by restricting outcomes to future states of affairs. Assuming CEM, however, these variants coincide with WO and FO, respectively.
See e.g. Parfit, p. 24, Scheffler(1), p. 1, and Williams(I), p. 87. MA entails AX, given the assumption that obligatoriness, Tightness, and wrongness are the only kinds of moral status. Most consequential-ists have presumably made this assumption.
Slote, Chapter III. He admits, however, that it is in line with present-day terminology to reserve the name ‘consequentialism’ exclusively for maximizing theories.
The locus classicus concerning the objection that consequentialism leads to violations of personal integrity is Willlams(l). In Slote, Chapter II, Slote argues that this objection should be put in terms of moral autonomy, rather than integrity.
Hurka, pp. 107–108. Hurka introduces the labels ‘absolute-level satisficing’ and ‘comparative satisficing’, which I have borrowed.
Adopting a somewhat different definition of comparative satisficing might make it possible to retain to-talism and still avoid this conclusion. One could claim that an action is right iff it realizes at least a certain percentage of the ‘greatest contribution to value’, that the agent is in a position to make. (Cf. Hurka, p. 108.) On a totalistic view of outcomes, however, it is far from clear how the ‘greatest possible contribution to value’ is to be defined. (Is it, for example, identical with the difference between the value of the outcome of the best alternative in the situation, and the value of the outcome of ‘doing nothing’? Or is it rather the difference between the value of the outcome of the best alternative, and the value of the outcome of the worst alternative? Or is it the difference between the value of the outcome of the best alternative, and the value of the outcome of what the agent would do if he were not to do the best alternative?) And even if this problem can be solved, this version of comparative satisficing is open to the objection made below. 29Admittedly, the description of the case is too meagre to say for sure what the common-sense moral judgement would be. But I believe that it would be easy to elaborate it (without changing the relative values of the outcomes), in such a way that common-sense morality would undoubtedly permit pressing B.
Hurka, p. 109.
’Sacrifice’ should here be understood in a wide sense, so as to include not only giving up something one already has, but also forgoing something one could get.
Besides, Slote would probably not be happy with a theory of this kind. He claims that there are actions which are supererogatory, according to common-sense morality, although they involve no sacrifice at all for the agent. (Slote, p. 47.) Hence, he is concerned with finding a satisficing theory that permits also (some) suboptimal actions which mean no sacrifice for the agent.
It is often claimed that there are situations where common-sense morality does not even permit maximizing the good. But this is a different objection, which I will not discuss here.
This, or at least a similar idea, is suggested by David Sosa. (Sosa, p. 116.) It also points in the direction of the ‘scalar consequentialism’ discussed by Slote. (Slote, Chapter V.) This theory, which Slote regards as an alternative to both maximizing and satisficing consequentialism, makes only (or almost only) comparative moral assessments of actions. Slote rebuts the possible objection that such a theory fails to be ‘action-guiding’, partly by pointing out that the information that a certain action is ‘morally better’ than every alternative, for example, is action-guiding for a person who is morally motivated. (A person who lacks moral motivation would not be moved even by the information that a certain action is morally ‘demanded’.)
Of course, a consequentialist holds that whether we ought to morally praise or blame somebody, is determined by the outcome of doing so. This belief may lead some consequentialists to reject the concepts of praise- and blameworthiness as meaningless or lacking in application. It might be more reasonable, however, to retain these concepts, but reject the general principle that a person ought to be praised (blamed) iff she is praiseworthy (blameworthy).
It must be admitted, though, that it is not entirely unproblematic to treat ‘obligatory’, ‘right’ and ‘wrong’ as synonymous to ‘uniquely morally optimal’, ‘morally optimal’ and ‘morally suboptimal’, respectively. For one thing, this usage may be uncommon and misleading. More seriously, as Wlodek Rabinowicz has pointed out to me, it might be objected that it obliterates important moral distinctions. Our common-sense concepts of obligatoriness, Tightness and wrongness seem to involve components (such as the concept of a demand or a prohibition) which cannot be explained in terms of moral optimality or suboptimality. The consequentialist who wishes to respond to the objections from supererogatoriness and moral autonomy in the way I have outlined, must explain why these components are not essential to morality. If he wishes to keep our common-sense notions of praise- and blameworthiness he may also get difficulties in explaining these concepts, since he cannot do this in terms of Tightness and wrongness.
See e.g. Sosa, pp. 107–108, Railton, p. 113, and Brink, p. 257.
On the other hand, if K is rejected, some form of probabilism would, I believe, be more plausible than factualism. If there are objective probabilities, and these are always knowable, objective probabilism would then seem preferable to subjective probabilism. If there are not even objective probabilities, or these are unknowable, subjective probabilism appears to be the best alternative.
For an explanation of the concept of expected value, or expected utility, see e.g. Gärdenfors & Sahlin, Chapter 1. (In decision theory, the value or utility to be maximized is usually taken to be the agent’s preference satisfaction.)
A version of objective probabilism is developed by Graham Oddie and Peter Milne. Their reason for preferring objective probabilism to factualism is that CEM may be false. (Oddie & Milne, p. 54.) In Oddie & Menzies, on the other hand, it is argued that objective probabilism is more plausible than factualism, even if CEM is true. (Oddie & Menzies, p. 516.) However, Oddie’s and Menzie’s argument seems to implicitly assume that K is false.
This objection to subjectivism is made e.g. by Fred Feldman (Feldman, pp. 46–48) and by David Sosa (Sosa, pp. 110–112). There are more moderate forms of subjectivism, which may partly avoid the present objection. One might suggest that the relevant subjective probabilities are not those which the agent actually has in the situation of choice, but rather those that she would have if she had the knowledge she ‘ought’ to have. But it seems extremely difficult to give a plausible criterion for determining what a person ‘ought’ to have known, in a certain situation. At the other end of the spectrum, there are extreme forms of subjectivism, according to which the subjective probability of an outcome should be weighted by the intrinsic value the agent believes this outcome to have, rather than by its actual value. Such extreme subjectivism seems to have no plausibility, however, since it may yield any hideous prescriptions whatever, provided that the agent’s beliefs about value are sufficiently bizarre.
Jackson(2), p. 471.
A problem for the factualist is to make plausible that we can have reasonable evidence that a given decision-making procedure satisfies the criterion. Otherwise, pointing to this criterion does not help to avoid the objection that factualism fails to be action-guiding. (Krister Bykvist made me aware of this problem.)
Jackson here refers to Peter Railton, who suggests roughly the same response as I have outlined above, to the objection that factualism provides no guide to action. (Railton, p. 117.) As far as I can see, however, nothing that Railton says indicates that he embraces the unacceptable decision-making procedure which Jackson describes in the quoted passage.
David Sosa proposes a different reply to Jackson’s example. Taking (subjective) risks is intrinsically bad, Sosa claims, wherefore it is reasonable to suppose that giving drug Y is actually right, according to factualism. (Sosa, p. 109.) I find it hard to believe, however, that it is intrinsically bad to take risks.
In the following I shall generally speak of ‘comparability of outcomes’, instead of ‘comparability of values of outcomes’, since the former mode of expression is less cumbersome.
A relation R is a weak ordering of the set A iff R is transitive in A, and for every x and y in A, either xRy or yRx. (Suppes, p. 222.)
If the possibility of pairwise incomparability of outcomes is allowed, there can be situations where there is no action such that its outcome is at least as good as that of every alternative. In such a case MA makes every alternative wrong. Since this conclusion seems rather contra-intuitive, the following amendment of the principle suggests itself: Ma’ An action ought to be performed (is right) iff its outcome is better than (at least as good as) that of every alternative, and an action is wrong iff its outcome is worse than that of some alternative. This modification excludes the possibility of situations where every alternative is wrong, but it can be disputed whether strengthening the criterion of wrongness is enough to handle situations with incomparable outcomes. Suppose the actions a, b and c are the only alternatives for some agent in some situation, and let A, B and C be the respective outcomes of these actions. Suppose further that both A and B are better than C, but that A and B are incomparable with each other. In this situation MA seems unsatisfactory, since it makes a and b wrong, as well as c. According to my normative intuitions, at least, only c is wrong in this case. MA’, on the other hand, implies that c is wrong, while a and b lack moral status (or have indeterminate moral status). This may be as it should; given the general idea that the moral status of an action wholly depends on facts about intrinsic value, it is natural that incomparabilities of value lead to normative indeterminacies. (This was suggested to me by Rabinowicz. If we assume a strong form of incompa-rability of value, according to which ‘better than’-statements sometimes lack truth-value, it would perhaps be even more appealing, on this view, to adopt a principle that also allows for strong normative indeterminacies, in the sense that statements of the form ‘a is right’ sometimes lack truth-value.) Against this conclusion it seems fairly reasonable to say that a and b are both right, since there is no alternative with a better outcome. (And since doing either of a or b ensures that a suboptimal outcome will be avoided. Cf. 6.3.) If one takes this position, MA’ may be amended, by weakening its criterion of Tightness, to: MA* An action ought to be performed iff its outcome is better than that of every alternative, an action is right iff there is no alternative with a better outcome, and an action is wrong iff it is not right. Note that MA* does not entail that two actions with incomparable outcomes always have the same moral status. If we change our example and assume that B and C as well as A and B are incomparable, while C is still worse than A, we get the result that b is right and c is wrong. This seems quite reasonable. (If A and B are incomparable in the stronger sense that the statements ‘A is better than B’ and ‘B is better than A lack truth-value, the statements ‘a is right’ and ‘b is right’ also lack truth-value, according to MA*. This is because it is neither true nor false, concerning a and b, that there is no alternative with a better outcome.)
On the distinction between a right-making characteristic and a decision-making procedure, see Bales(l).
Williams claims that “[t]he fathers of utilitarianism [the most well-known consequentialist theory] thought of it principally as a system of social and political decision, as offering a criterion and basis of judgement for legislators and administrators”. (Williams(I), p. 135.)
This does not necessarily mean that C is applicable to all performable actions. Bergström has argued that a ‘normatively reasonable’ consequentialist theory only applies to actions that are members of ‘relevant alternative-sets’. (See 6.1.) If I have understood him rightly, he holds, however, that there is a relevant alternative-set for every agent in every situation of choice. (See Bergström(l), Chapter 2, Section 4.)
R.M. Hare argues, with some qualifications, that overridingness is a part of the ‘logic’ of the moral terms. (Hare, Chapter 3.)
This seems to be in line with most descriptions of consequentialism. See e.g. Parfit, p. 24, Schef-Fler(l), p. 1, Kagan, p. xi, and Williams(I), pp. 87–89.
See e.g. Parfit, p. 27, Scheffler(2), p. 5, Sen, pp. 205–206, österberg, p. 129, Rabinowicz(2), p. 19, Sosa, p. 114, Oddie & Milne, p. 42, and Vallentyne(I), p. 28. (Vallentyne uses the terms ‘agent-insensitive’ and ‘agent-sensitive’ instead of ‘agent-neutral’ and ‘agent-relative’.)
Rabinowicz claims, in effect, that TMA-moralities which assume FO are not quite agent-neutral, since they give agents acting at different times different ‘maximanda’. (Rabinowicz(2), p. 21.) (If P acts later than P’, the future state of P’s action is a proper part of the future state of P“s action.) TMA-moralities that assume WO, on the other hand, are wholly agent-neutral, according to Rabinowicz’s definition of agent neutrality.
Cf. Sosa, p. 115.
See e.g. Scheffler(2), p. 5, and Parfit, p. 27.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1995 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Carlson, E. (1995). Consequentialism Characterized. In: Consequentialism Reconsidered. Theory and Decision Library, vol 20. Springer, Dordrecht. https://doi.org/10.1007/978-94-015-8553-8_2
Download citation
DOI: https://doi.org/10.1007/978-94-015-8553-8_2
Publisher Name: Springer, Dordrecht
Print ISBN: 978-90-481-4571-3
Online ISBN: 978-94-015-8553-8
eBook Packages: Springer Book Archive