Hostname: page-component-78c5997874-94fs2 Total loading time: 0 Render date: 2024-11-13T11:31:01.625Z Has data issue: false hasContentIssue false
  • English
  • Français

Utilitarianism and Cooperation*

Published online by Cambridge University Press:  13 April 2010

Jordan Howard Sobel
Jordan Howard Sobel
Affiliation:
University of Toronto
Get access Rights & Permissions [Opens in a new window]

Extract

Donald Regan's book is a study of forms of utilitarian theory and of properties to which these forms may aspire. He deals exclusively with “fully objective” theories that seek to systematize and set conditions for obligations that an agent is supposed to lie under as a consequence of features of situations in which he finds himself, with no dependence on his possibly defective beliefs, attitudes, and values (except as features of his situations). Metaethical questions concerning the nature of these obligations are not raised. And practical problems that could be encountered in attempts to apply these theories receive very little attention. They are said to lie outside the main project of the book which is to explain and then to circumvent a certain apparent logical impasse that utilitarianisms ca. 1980 are supposed to confront. The book consists for the most part of abstract, often very long and difficult, arguments, and detailed elaborations of a small number of minimally specified, highly artificial cases of the sort now common in semi-formal utilitarian theory. In the present note, while leaving unquestioned and unstated the presuppositions of Regan's book, I first outline its central two-stage “argument”. Next I explain several distinctions important to issues it addresses and relate these distinctions to several of Regan's theses. I then discuss prima facie cases against his main claim, according to which the apparent impasse that utilitarianisms confront is circumvented by a new utilitarian theory designed for the task.

Type
Critical Notices/Etudes critiques
Information
Dialogue: Canadian Philosophical Review / Revue canadienne de philosophie , Volume 24 , Issue 1 , Spring 1985 , pp. 137 - 152
Copyright
Copyright © Canadian Philosophical Association 1985

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 Consulting the Index one learns that these words provide only a “summary statement” of CU, and that a “spelled-out” version can be found on 165–166. (The “spelled-out version” is one page long.) Also of interest are the “basic idea” of CU succinctly stated in the Preface on x, “what CU contemplates” which is explained and summarized on 135–136, and the “idea implicit in CU” that is set out in six lines on 166–167.

2 Regan does not give a “precise and completely general” definition of exclusive act-orientation (109). His strategy is to work with an intuitive notion of this condition, and a case in which the implications of this intuitive notion are clear (106–109). A difficulty with this strategy is that it makes problematic whether or not all pre-Regan utilitarianisms are exclusively act-oriented, and so makes problematic Regan's claim that CU not only succeeds but succeeds where all previous theories have failed. Frey, Raymond, in his review of Regan's book (Mind 92 [1983], 297–298),Google Scholar makes this point with particular reference to the two-level utilitarian theory that R. M. Hare advocates in Moral Thinking. Of course, even if (as seems true) not all pre-Regan utilitarianisms are “exclusively act-oriented”, though Regan's argument (113–116) would not show that no previous theory was both “individually” and “collectively efficient”, this could be true. A way to show that this was true would be to identify and set down in turn the presumably few pre–Regan utilitarianisms that do not clearly exhibit exclusive act-orientation, and that Regan's general argument misses. One might start with Hare's two-level theory (even though it is, if we go by publication dates, not“pre-Regan”), and observe that it has neither PropAU nor PropCOP. Hare's theory does not aspire to these properties.

3 CU is held to have not only PropAU and PropCOP, but the property of “adaptability”, which Regan defines as follows: “A theory T is adaptable ifand only if, in any situation involving choices by any number of agents, the agents who satisfy T in that situation produce by their acts taken together the best consequences they can possibly produce by any pattern of behaviour, given the behaviour of agents who do not satisfy T” (6). Adaptability is supposed to entail, but not to be entailed by, the conjunction of PropAU and PropCOP.

4 It is nearly certain that Regan intends PropCOP in a subjunctive sense. Immediately following its (mood-ambiguous) official statement he writes, “In other words, a theory T has PropCOP ifand only if universal satisfaction ofT (satisfaction by all agents all the time) would guarantee the best consequences that any pattern of behaviour by the universe of agents could possibly produce” (5, italices mine). That PropCOP is almost certainly intended in a subjunctive sense is strong evidence that the same is true of PropAU, and ofthe property ofadaptability which is supposed to entail the conjunction of PropAU and PropCOP.

5 It is clear that the words “the consequences of everyone's satisfying a theory T would be best” would here be intended by me in a collective sense. (I discussed the collective/distributive ambiguity of such forms in “‘Everyone’, Consequences, and Generalization Arguments”, Inquiry 10 [1967], and proposed a notation in which it corresponds to a difference in the relative scopes of an operator and a universal quantifier. Adapting the proposal to the present form, its collective sense could be symbolized by “β [x] Tx” and its distributive sense by “[x] β Tx”.) And the words “the consequences of everyone's satisfying a theory T” are of course intended in a sense that contrasts with the most natural sense for “the consequences of everyone's doing what in the circumstances (including those concerning what persons will do, and would do if other persons behaved differently than these others will behave) would satisfy a theory T”. Similarly remarks are in order concerning the intended sense of ”the consequences of everyone's doing what in the circumstances of the case would satisfy T“.

6 It can seem that Regan has by implication objected in advance to equating the pattern {R2, C2} with “everyone's doing what in the circumstances of the case AU requires” and with “everyone's doing what in the circumstances of the case would satisfy AU” (57–58). But properly interpreted (and it should now be obvious that even these formulae are ambiguous and open to substantially different interpretations) these equations are correct. For in the case it is given that each would do his first action whatever the other did; this is given as one of the circumstances of the case. And so it is clear that AU requires in the circumstances of the case that Row do R2 and that Column do C2, and that everyone's doing what AU requires in the circumstances of the case is the pattern {R2, C2}, as is everyone's doing what in the circumstances of the case would satisfy AU. It is no argument against these identities that if everyone were to do what AU requires of him in the circumstances and what in these circumstances would satisfy AU, then no one would be doing what AU required of him or what would (then, in those circumstances) satisfy AU. Regan distinguishes “adaptability”' as defined in a previous notefromwhat can be termed “general adaptability”. A formula for the latter comes from that for “adaptability” by putting “any agents” for “the agents”. He argues that these properties are equivalent (108). Indicative versions of these properties do seem to be equivalent. But subjunctive versions are not. Combining the distinction Regan makes with the kind that separates PropCOP(s)-1 from PropCOP(s)-2 leads to four subjunctive forms of adaptability that are independent of one another.

7 My argument to show that no theory of any sort can have both PropAU(s) and PropCOP(s)-2 might be compared with Regan's argument to show that no “exclusively act-oriented theory” can have both PropAU and PropCOP (114–116).

8 Indeed, CU seems to be, and is I think intended to be (though Regan is not explicit on this matter), extensionally equivalent with AU with respect to actions that would satisfy AU and that could be involved in the satisfaction of CU. Of course AU and CU are not unrestrictedly equivalent with respect to what would in cases satisfy them, for satisfying AU in a case will involve only some action, whereas satisfying CU will at least typically involve not only an action but certain beliefs and attitudes as well. “But how could AU and CU be extensionally equivalent with regard to actions? They are so different. AU is, in particular, –indeterminate— in a sense in which CU is not.” AU is “indeterminate” in the sense that there are cases in which it “can be universally satisfied by different patterns of behaviour which produce consequences of different value”, and I believe that CU is not “indeterminate” in this sense (20). But it does not follow that there is a case in which, for some agent, the set of acts that would satisfy AU is different from the set that could be involved in the satisfaction of CU. It does not follow that AU and CU are that different, and I think that they are not supposed to be that different. I think they are supposed to be extensionally equivalent with regard to actions. A principle could have PropAU without being extensionally equivalent with AU with regard to actions: consider versions of AU that include “tie-breaking” clauses. CU does not have to be extensionally equivalent with AU with regard to actions just because CU has PropAU (though this property does entail such equivalence for “no AU-ties” cases). But I think that CU is supposed to be equivalent with AU in this way, and I think it is equivalent with AU in this way, though given the complexity of CU it would be difficult to demonstrate this equivalence. To convince yourself of this equivalence, I recommend that you try to spell out a case in which AU and CU diverge with respect to actions that could be involved in their satisfactions. The intended relation between AU and CU is I think this. AU and CU are supposed t o be extensionally equivalent in a certain distributive sense. For any agent in any situation, the acts that would satisfy AU are supposed to be identical with the acts that could be involved in his satisfaction of CU. But AU and CU are not extensionally equivalent in a collective sense. It is not that case that in any situation, all sets of actions that could constitute joint satisfactions of AU by all agents are identical with sets of actions that could be involved in everyone'sjointly satisfying CU. We can say that CU is supposed to have PropAU-DisEqttiv, though CU certainly does not have PropAU-CollEquiv.

9 Let me offer a non-minimal statement of Regan's principal thesis. We can say that a principle has PropCOP-CollEquiv when it is with respect to actions that would be involved in satisfactions extensionally equivalent in a collective sense with the principle of coordinated optimization, “COP An act is right if and only if it is prescribed (for the agent whose act is in question) by that universal prescription for action, the universal satisfaction of which would produce the best possible consequences.” PropCOP-CollEquiv is to be understood on the model of PropAU-CollEquiv. (See note 8 above.) For a non-minimal statement of Regan's main claim for C U we have that CU has both PropAU-DisEquiv and PropCOP-CollEquiv. CU, it is claimed, is with respect to actions that would be involved in satisfactions, extensionally equivalent with AU in a distributive sense, and with COP in a collective sense. Since in contrast with these principles, CU is not exclusively act-oriented, it is neither distributively nor collectively extensionally equivalent with respect to all aspects of satisfactions with either AU or COP.

10 Everyone in a 2(selfish)-world does the action he would do were he to satisfy CU. but no one in such a world satisfies CU, for no one in such a world is initially willing to cooperate with all those with whom, were they willing, he could cooperate. And “the first requirement of CU is precisely that each agent –hold himself ready—” (137): “ready t o do his part in the best pattern of behaviour for the group of co-operators, whoever precisely the other members of that group turn out to be” (135).

11 “But how could adherents of CU come to have such unfortunately convergent true beliefs about each other?” The answer is. In whatever ways they could come to have fortunately convergent true beliefs in a case (such as the one under discussion) that involves multiple unequal utilitarian equilibria. A little bird could tell them. They could guess and be right. They could tell each other of their beliefs and intentions and believe what they hear. And so on.

12 AU is not only an “exclusively act-oriented” theory, but an “exclusively individual-oriented” one: AU addresses itself only to individuals; demands by it on “unincorporated groups” are entirely derivative. AU of course has PropAU(s), and lacks PropCOP(s)-l. CU, in virtue of its non-exclusively mr-oriented character, joins PropCOP(s)-! with PropAU(s). It is much easierto combine these properties in a theory that departs from AU in the other way and is non-exclusively individual-oriented. More precisely, it is easier to combine in a non-exclusively individual-oriented theory versions of these properties reached by counting as “agents” not only all individual agents (including incorporated groups of individual agents), but also all groups of individual agents (including “unincorporated” ones). Earl Conee describes a theory that does just this. He sets out this theory, which imposes AU-type obligations not only on individuals but on all groups of individuals, in his review of Utilitarianism and Cooperation (Journal of Philosophy 80 [1983])Google Scholar.

13 Also relevant here is the fact, if it is a fact, that AU and CU are extensionally equivalent with respect to actions. See note 8 above. If AU and CU are extensionally equivalent with respect to actions, then CU is not recommended over AU on the ground that it solves problems, for example, “coordination problems”, that embarrass AU. But even if CU is extensionally equivalent with AU in a distributive sense; speaking and for what this is worth, CU might recommend itself to a utilitarian God as a law for His subjects, if as seems true CU is extensionally equivalent in a collective sense with COP. Questions 1 leave open are first, whether or not CU would recommend itself to a utilitarian God as a law for His subjects everything considered including problems with and costs of implementations of CU by His subjects, and second, if CU would recommend itself in this way, whether or not this figurative fact is relevant to any “runnings” of proper interest to us, and if so, how.

14 “But is AU a better approximation to a correct theory of moral choice? Is it a better approximation to a correct theory of objectively right moral choice?” Admittedly, these are questions of interest to some theorists, and may be questions regarding AU of interest to Regan. They are, however, questions upon which I feel compelled to pass. My problem is not so much that I do not understand “objective moral Tightness”, as that to the extent that I do understand these words in their ordinary ways I do not believe in objective moral rightness. I share J. L. Mackie's moral scepticism.

15 “Cooperative expected-subjective-utility-maximization” models of rational choice such as David Gauthier's, in their attempts to minimize sacrifices of would-be cooperators to non-cooperators while maximizing cooperation between would-be cooperators, are threatened by “circularities” similar to those that threaten CU, and display propensities to similar degrees if not modes of complexity.