The Ideal Agent Theory
Philosophers familiar with G. E. Moore's "naturalistic fallacy argument" have often doubted its value in establishing Moore's most celebrated claim, namely, that the "fact-value" distinction precludes the possibility of what might appropriately be termed a "natural science of ethics." 1 Recent work in the philosophy of science has provided additional reasons for concluding that Moore's argument, thus construed, is unsound. 2 These developments suggest that a new look at the entire range of issues facing the program of ethical naturalism is overdue. Moore's "fallacy" is not the only problem facing ethical naturalism. But the prospects for a true science of ethics are markedly improved by the decisive refutation of an argument that has seemed to more than a generation of twentieth-century philosophers the coup de grace for naturalism in ethics.
In this paper I support the plausibility of the naturalists' program by presenting a theory that I claim is both a moral theory and a natural theory capable of empirical corroboration. This theory takes the form of a theory of free action, thus drawing upon intuitions as modern as Kant and as ancient as Plato linking the issue of right action with the issue of free action. However, our theory is unusual in that it links the norms of right action to the natural norms of lawlike free action. We are thus, on the issue of free-will, compatibilist to the core, finding free action not only compatible with, but dependent on, its own lawlike character. We thus support a theory of right action devoid of the usual metaphysical, or transcendental, elements. I shall label this theory The Ideal Agent Theory because of its affinities to Roderick Firth's "Ideal Observer" analysis of absolutist ethical statements. 3
Such a theory of "lawlike freedom" must overcome many familiar obstacles. On the issue of free-will, I refer the reader to my earlier papers outlining my flavor of free-will compatibilism and the underlying view of "supervenient" time upon which it rests. 4 At the very outset, of course, I must elaborate the position that I believe can be maintained in response to Moore's celebrated naturalistic fallacy argument. Moore surely intended his requirement of analyticity for ethical definitions as a criterion of explanatory relevance, as reflected in his motto, "everything is what it is and not another thing" (Frankena, 57-8). Thus, in refuting Moore, I must provide a viable alternative criterion of explanatory relevance for naturalistic ethical theories. I intend to show that such an alternative criterion of explanatory relevance can indeed be found with the help of recent work in the philosophy of science. This focus on the philosophy of science will serve to get at what I take to be the real source of the disenchantment with ethical naturalism which is evident in Moore and continues to characterize contemporary moral philosophy at a time when the arguments of Moore are no longer persuasive for many. The real source of this disenchantment is the suspicion that there is just not a sufficient consensus regarding moral facts for a naturalistic ethics ever to get off the ground, so to speak. It can seem that, because of this lack of pre-scientific consensus on the domain of moral "explananda," the most we could ever hope to achieve through science would be competing "psychologies" of morality, never a truly accepted science of morality. This lack of a pre-scientific consensus on the domain of moral facts makes it appear that the criteria for deciding between competing naturalistic moral theories will remain, as has always been the case with moral theories, a decision based wholly on moral criteria. However, I shall show that this assumption is incorrect, not only with respect to the irrelevance of non-moral criteria in the justification of a moral theory, but also with respect to the idea that moral criteria are inevitably subjective and thus themselves useless in the justification of a science of ethics.
Obviously we need a clearer picture of the role of criteria in the justification of a science of ethics. I shall elaborate two kinds of criteria based examples from the sciences. The first is the criterion of explanatory relevance (ER). A science of the morally good must have some discernible relevance to a pre-scientific domain of moral facts. But, explanatory relevance in science has little to do with linguistic entailment relations tying theory to pre-theoretic facts. The most we can normally require is that the scientific explanation provide verifiable theoretical facts in a relationship of rough co-extensionality with pre-theoretic facts. Moreover, ER is not the only criterion for a science of ethics. Examples from the natural sciences suggest that often the deciding factor in the adoption of one theory from a multitude of competing theories is the independent corroborability of the chosen theory. This kind of "consilience" of scientific theories has been advocated by Edward O. Wilson as a metaphysical issue having to do with the unity of scientific knowledge (1998, 8-9). However, the key advantage of the consilience Wilson describes appears to be the added epistemic support provided by the independent corroboration of the consilient theory. I shall thus refer to this second criterion as the requirement of "Independent Epistemic Support" (IES). Because of the need for IES, it could very well be that the deciding factor in the adoption of a naturalistic ethical theory might be its consilience with non-moral domains of natural facts.
Once the criteria of ER and IES are established, we can proceed to the general justification of the idea of ethics as a science of free action. Ethical theories of this general type can be defended by showing their potentially superior ER and IES in the explanation of moral facts. For example, we can show the potential superiority of natural theories over non-natural theories of ethical facts with respect to IES. Secondly, with respect to ER, we can show the potential superiority of naturalistic theories of free action over naturalistic theories of the consequentialist type in explaining certain peculiarities of ethical or value attributes noted by G. E. Moore, namely, the "organic wholes" comprising certain ethical values that seemed to Moore and many non-naturalists since Moore as a problem for consequentialist theories.
Finally, I shall consider two problems which have appeared in the past to prevent the explanation of ethical facts in terms of a naturalistic theory of right action. The first is the problem of Free-Will, which I have suggested can be resolved through a compatibilist argument based in part on a view of the "supervenience" of time, that is, time as supervening on change. The second problem concerns the moral relativism seemingly entailed by any explanation of right action in terms of free action. This way of thinking about free action reflects an ancient dogma, a true folk paradigm whose historical roots are at least as old as Aristotle, and whose influence can be seen in natural theories of human action up to the present time. Alan Donagan describes quite clearly Aristotle's historical role in the development of this paradigm.
The propositional attitude by which Socrates explained his remaining in prison was a belief (doxa): the belief that to do so would be for the best. Aristotle ventured a correction. Socrates remained in prison because he chose to; but he "can hardly be said" to have believed to... Since no belief is necessarily acted on, actions cannot be explained solely by their agents' beliefs. Propositional attitudes of some other kind must be at work: attitudes the natural behavioral expression of which is trying to bring about... Aristotle's generic term for such attitudes was "orexis," which the medieval Aristotelians rendered into Latin as "appetitus." 5
In spite of the pervasive influence of this "Aristotelian" paradigm, it is in fact quite vulnerable to criticism according to the criteria of ER and IES. (As a part of our "folk psychology" it shares the same weaknesses of other folk elements of the pre-scientific conception of the self.) I shall propose an alternative theory that questions the separability of the human willing faculty from the human cognitive faculty, that is, the traditional separation of beliefs from desires. If we take desires to be integral with complex structures of human beliefs, then the way is opened for a natural theory of free agency in which the natural norms of human beliefs might correlate with natural norms of actions resulting in a lawlike way from those beliefs. This in turn results in the possibility of a non-relativist theory of right action based on the hypothesized free actions of an agent who is in some non-question-begging sense an ideal observer in the manner envisioned by Roderick Firth. If desires are integral with beliefs, then the ideal observer, if free to act, becomes at the same time an ideal agent.
I conclude with a description of the Ideal Agent, thus conceived. A crucial advantage of the Ideal Agent theory is that it connects in a necessary way the "public" nature of beliefs with certain essential elements of moral actions, namely, the attributes of fairness, disinterestedness, and altruism, that recent thinkers have thought essential to rational, and moral, "reasons" for actions. 6 A second important advantage is that the integral connection of belief structures to free actions also provides an explanation of the compelling, or action requiring, nature of moral discourse. The Ideal Agent, or any mature free agent of reasonable knowledge must find moral "reasons," if true, compelling. They are, after all, the basis, in the relevant action context, for the free actions of such a free agent.
Could the Ideal Agent theory, or some variant, truly be a Moral Science? The idea that such a naturalistic theory might be true has some interesting consequences for the domain of ethical discourse. First, the Ideal Agent theory presents a picture of moral facts depending on, and explained by, natural facts of human psychology in relation to the current world context. If moral facts supervene on natural facts in the sense employed by Hare, it is only in the way of this kind of weak supervenience. "Martians" might have an "ethics" hardly recognizable by humans as meriting the term. However, we would probably want to extend the term "ethics" only to those species capable of objective thought. For without this kind of objective thought it's hard to conceive of such a species as conscious, or capable of intentionality in the appropriate sense. And in light of the role of public language in fostering this capability for objective thought, we might hypothesize that any species with an "ethics" could be surmised to be a "social" species in some sense of the term.
There are other important limitations of this Ideal Agent theory as an ultimate theory of value. First, even if we assume the scientific dictum that "similar causes have similar effects," we can expect no more than a "quasi-absolutism" with respect to our theory of moral, or right, action. Although there are many reasons for this, the most telling is the point made by Quine on the indeterminacy of translation. Quine's thesis suggests, as a corollary, that even if we could isolate in a laboratory the disembodied beliefs of agents with "similar beliefs" we should not expect to find either identical "internal representations" in the sense of the traditional empiricist theory of "ideas," or, in the picture of beliefs favored by modern functionalists, identical brain states. What people really mean or believe when they "believe the same thing" loses significance when we take that issue beyond the public behaviors that go to make up the rules of grammar and semantics for such statements as "X believes that p." On the other hand, we perhaps ought not exaggerate the potential imperfections in the absolutism that results from the Ideal Agent theory. After all, our knowledge that members of a language community have certain beliefs is entirely bound up in their public behaviors. And, as we have seen, part of the behavior supporting these linguistic norms is morally relevant behavior. If a person in a language community acts in a way that ignores morally relevant contexts, one of our responses will be to question that person's beliefs.
Another final limitation is that an Ideal Observer theory, of which the Ideal Agent theory is a variation, presupposes the cognitive ability of the Observer or Agent to seek and recognize the truth. And in an age of Kuhn, Feyerabend, and Rorty, this ability to "seek the truth" has become much more mysterious than it may have appeared to thinkers early in this century (Carnap, Hempel, and Ernest Nagel come to mind) when perception and logic were all that seemed to be required. It must be admitted by any epistemologist today that the reform and optimization of our structures of belief is shot through and through with issues of "value", albeit of an epistemic variety. We may indeed find, in the Ideal Agent theory a natural science that purports to be an ethical theory. But scientific theories are always "defeasible" in a sense that will remind any moralist familiar with Principia Ethica of Moore's own "open question test." A natural science of ethics will always face the neo-Moorean challenge "is this theory really acceptable as a theory of good?" And, as we have seen, although this is not a question to be answered purely on "moral" grounds, it is still always an intelligible question, as it is with any scientific theory.
1 See, for example, Philippa Foot, ed., Theories of Ethics, Oxford, Oxford Univ. Press, 1967, pp. 1-13, and the included article by W. K. Frankena, "The Naturalistic Fallacy," pp. 50-61.
2 See for example, Lawrence Sklar (1993,114-16) on "scientific identificatory definitions", and W. V. Quine (1992, p. 8) on "proxy functions."
3 Roderick Firth, (1952, 317-345) "Ethical Absolutism and the Ideal Observer," Philosophy and Phenomenological Research, XII.