Is God Good By Definition? (1992)
[This article was originally published as “Is God Good By Definition?”, Religious Studies, 28, 1992, pp.467-474.]
As a matter of historical fact, most philosophers and theologians who have defended traditional theistic views have been moral realists. Some “divine command” theorists have held that the good is constituted by the content of divine approval–i.e. that things are good because, and insofar as, they have divine approval. However, even amongst those theists who hold that the good is independently constituted–i.e. those who hold that God’s pattern of approval is explained by the fact that he approves of all and only that which is good–the dominant meta-ethic has been strongly realistic.
As a further matter of historical fact, one of the main motives for the development of non-realist meta-ethics has been the desire to give an adequate atheistic account of the nature of the good. Thus many subjectivist, projectivist, and error-theoretic accounts of the good were developed in the context of atheistic enquiries. Of course, atheistic accounts of the good can be realist–e.g. G. E. Moore’s non-natural objectivism. However, the question which I wish to take up is whether theistic accounts of the good can be non-realist.
In fact, I shall argue that the traditional philosophical conception of God requires a commitment to moral realism. Despite the range of non-realist meta-ethics which have been developed, there is none which is compatible with theism. Consequently, there is a hitherto ignored argument against theism which emerges for assessment, viz:
1. Ethical realism is a necessary consequence of traditional theism.
2. Ethical realism is false.
3. (Therefore) Traditional theism is false.
In this paper I shall focus on 1.–i.e. I shall not attempt to provide an argument for 2. However, I believe that there are independent arguments–i.e. arguments independent of the debate about theism–which favour the view that ethical realism is false. Consequently, I believe that the argument of this paper makes a contribution to the case against theism.
What I take to be the traditional conception of God has been expressed by Richard Swinburne. In The Existence Of God (Oxford University Press, 1979), he writes:
I take the proposition “God exists” … to be logically equivalent to “there exists a person without a body .. who is eternal, is perfectly free, omnipotent, omniscient, perfectly good, and the creator of all things”. I use “God” as the name of the person picked out by this description. (p.8)
I do not think that one ought to say that “God exists” is logically equivalent to “there exists a person without a body… etc”. For, in saying this, one is committed to the view that if, for example, (i) there exists a person without a body who is eternal, perfectly free, omnipotent, omniscient, very (but not quite perfectly) good, and the creator of all things; but (ii) there is no person without a body who is eternal, perfectly free, omnipotent, omniscient, perfectly good, and the creator of all things, then God does not exist. This seems to me to be an odd view to take; in the circumstances described, it seems to me that it would be more natural to say that God does exist, but that he is not quite as we imagined him to be.
However, this is not a serious problem. What I think Swinburne ought to say is that we use the expression “person without a body… etc” to fix the referent of the expression “God”: God is the being, if any, which comes closest to satisfying this description. If two or more beings satisfy the description equally well, then–strictly speaking–God does not exist (though perhaps certain gods do exist). Moreover, given this amendment to Swinburne’s account, it then becomes a substantive theory that God is a “person without a body…etc”.
I shall assume without further argument that the amended account does capture the traditional conception of God. The crucial point for what follows is that many theists have been committed to the view that it is a definitional or constitutive or necessary truth about God that he is perfectly good–and it is this view which I claim is incompatible with a non-realist meta-ethic.
Before I turn to the details of the argument, I wish to introduce an independent reason for interest in it. This reason arises in the context of a dispute between Michael Tooley and Richard Swinburne over the correct meta-ethical background in which to situate a discussion of the problem of evil.
In “Does Theism Need A Theodicy?” (Canadian Journal Of Philosophy 18, 1988, pp.287-312), Swinburne writes:
I need to establish a position on the status of moral judgements. I hold that they have truth-value; some are true and some are false. … There could only arise an issue as to whether certain evils were compatible with the existence of a good God if goodness and evil were properties which belonged to persons, actions, and states of affairs, and judgements which affirmed or denied their existence had a truth value. (p.290)
In response to this passage, Michael Tooley (in “The Argument From Evil”, forthcoming) comments that:
Embedded in what is otherwise a very thoughtful discussion, these claims by Swinburne that the problem of evil arises only if moral judgements are either true or false, and only if goodness and evil are real properties of persons, actions and states of affairs, are rather jarring. For … suppose that John Mackie’s error theory is correct, and that, although we are ascribing non-natural properties to actions when we describe them as right or wrong, and to states of affairs when we describe them as good or bad, the world in fact contains no such properties. All of our “positive” ethical beliefs would be false in that case, but that would not be a barrier to some of John’s ethical belief being logically inconsistent with some of Mary’s, nor to some of John’s ethical beliefs being mutually inconsistent, or to their giving rise to inconsistencies when combined with some of his non-moral beliefs about the world. (p.8)
Tooley’s point is that questions about the consistency of sets of ethical beliefs can be raised quite independently of considerations about the ontological status of the objects of those beliefs–and, from this, it might seem to follow immediately that “the problem of evil” can also be investigated quite independently of considerations about the ontological status of the objects of moral judgements.
However, I think that Tooley has been mislead by Swinburne’s (apparent?) suggestion that any dispute about the compatibility of the existence of evil with the existence of certain non-moral states of affairs requires the background assumption of moral realism. Given that Swinburne holds that it is a definitional property of God that he is morally perfect, it follows immediately that a being which lacked the attribute of moral perfection could not possibly be God. Consequently, it would be inconsistent for Swinburne to adopt Mackie’s error theory: if there is no such property as “moral perfection”–or if there is such a property, but it is the case that nothing lies in its extension–then, according to Swinburne’s definition, God does not exist.
Now, of course, this argument does not show that Swinburne’s definition of God commits him to the meta-ethical view that moral judgements have truth values (and that goodness and evil are properties which belong to persons, actions, and states of affairs). However, as I shall now go on to argue, I think that it is plausible to suggest that Swinburne is (perhaps unwittingly) correct when he suggests that the problem of evil only arises in the context of the assumption of meta-ethical realism.
There are so many non-realist accounts of the nature of moral properties that I cannot possibly hope to discuss them all. However, I think that the theories which I shall consider are representative. My aim is to make it appear plausible that there really are no non-realist accounts of the nature of moral properties available to traditional theists.
1. Straightforward Subjectivism: First, let us consider the simple view that when a speaker says “x is morally perfect” what that speaker does is to express a maximal pro-attitude towards x. In general, the idea behind this view is that value judgements–“x is good”, “x is admirable”–are expressions of pro-attitudes. In addition, the view also holds that the proper use of expressions such as “x is morally perfect” is to express maximal pro-attitudes about x–i.e. pro-attitudes of maximal intensity. (With minor changes, the following discussion should apply to all theories which hold that moral judgements are merely subjective expressions of attitudes or emotions.)
One difficulty which this view presents, to the theistic project of definition, is that there is an indexical element in the use of the expression “x is morally perfect”: different objects will evince ejaculations of this expression from the mouths of different speakers. How, then, could such an expression be any use in the community-wide definition of a single object? This would seem to make as much sense as the supposition that the expression “object nearest to me” defines the same speaker-independent object no matter from which mouth it issues. Of course, this difficulty would be avoided if we all approved of the same things to the same extent. But we don’t; there are many different patterns of approval and disapproval which are exhibited, even by monotheists.
However, there is also a more important difficulty. If I adopt this sort of subjectivist emotivist meta-ethic, then I recognise that it is quite appropriate for one person to say that an object (action, thought, etc.) is morally good even though another person says that the same object is morally bad. Moreover, I recognise that there is no good reason to suppose that there will be universal convergence in the application of the expression “morally perfect”. Consequently, I recognise that it might happen that the theistic definition picks out a unique object in some mouths, but not in others. (Of course, it might also happen that the definition picks out no object in any mouth.) Should I subscribe to the theistic definition? Well, what reason do I have for thinking that my use of the expression “morally perfect” is one of the privileged ones which does in fact lead the thesitic definition to pick out a unique being? Surely I should be guilty of a very unchristian hubris to think that I am especially privileged in this respect. (Since even monotheists disagree in their moral judgements, the fact that I am a particular sort of believer cannot be enough!)
The conclusion which I draw from this is that if one subscribes to the theisitic definition, then one should not also adopt a straightforward subjectivist meta-ethic.
2. (Blackburn’s) Projectivism: A more sophisticated subjectivist meta-ethic has recently been defended by Simon Blackburn (see, e.g., Spreading The Word, Oxford University Press, 1984–chapter 6). On Blackburn’s projectivist view–as on the straightforward subjectivist view which we considered above–moral judgements are a species of non-descriptive judgements which serve to express our desires (and other non-cognitive attitudes). However, Blackburn’s theory is also quasi-realistic: i.e. it attempts to ground our ostensibly realist moral talk and moral practice within a more broadly projectivist framework. Thus, for example, in “Just Causes” (Philosophical Studies, vol 63, 1991, pp.3-17), Blackburn attempts to show that projectivism can account for the intuition that there is a common and communicable topic involved in any moral debate.
Blackburn’s suggestion is that the projectivist can say that, in any worthwhile moral debate, the participants will recognise that there is some range of features to which it is not wildly idiosyncratic to respond with some among a range of responses. Then, the common subject matter of the debate is constituted jointly by the range of features and the range of responses. (E.g: Socrates and Thrasymachus are both talking about justice because they are both responding to societal structures with attitudes of favour and disfavour.)
Perhaps the success of this response can be granted to the projectivist; but it is little help to the theistic project of definition. Suppose we agree that we shall hold that, by definition, God is morally perfect. Suppose further that we agree that we all mean by this that God possesses a set of properties which fall inside the limits imposed by idiosyncracy on what can count as “moral perfection”. Then the following will become a perfectly intelligible position: I believe that God exists, I believe that God is morally perfect, and yet I do not have the slightest disposition to regard God favourably. (After all, for all I know, he may be utterly monstrous–e.g. he may agree with Thrasymachus about justice!)
Now, I grant that this position is intelligible. However, I also hold that it is one which no traditional theist would wish to adopt. Part of the traditional view is that God’s moral perfection helps to make him a suitable object for universal worship and adoration. But, on Blackburn’s projectivist reconstruction of the notion of moral perfection, that part of the view is bound to be lost. If God is to be a suitable object for universal worship and adoration, then surely it must turn out not only that he is perfectly good according to the lights of each, but also that (according to each) he is perfectly good in a sense which (ceteris paribus) can motivate each to be disposed to regard him favourably.
So I conclude that (Blackburn’s) projectivism cannot provide a notion of moral perfection which is suitable for the theistic project of definition.
3. Response-Dependent Accounts: We can see the (above) highlighted feature of Blackburn’s projectivism as an attempt to reduce the intersubjective variation in the content of moral concepts which is found in straightforward subjectivism. The final theory which I wish to consider–the view that evaluative moral concepts are response-dependent–can be viewed as a further attempt in this same direction.
On a response-dependent view, it is an a priori truth about moral concepts that they satisfy the following constraint: objects fall under these concepts iff those objects are such as to produce certain non-trivially specifiable responses in certain non-trivially specifiable subjects under certain non-trivially specifiable conditions. So, for example, if moral goodness is a response-dependent concept, then it will be the case that an object is morally good iff it is such as to produce responses R in subjects S under conditions C. (Cf. Mark Johnston “Dispositional Theories Of Value”, PAS, 1990 for further characterisation and discussion.)
Now, in one sense, there is no barrier to the supposition that a response-dependent concept might be used in the definition of an object. Suppose–for the sake of an example–that colour concepts are response-dependent, so that an object is red iff it is such as to produce certain responses (looking red) in certain subjects (not colourblind, mentally alert, etc) under certain conditions (good lighting, etc). Consider the description: “the largest red object in the universe”. It is clear that this description does succeed in picking out an object–at least if the universe is finite and there is not a tie for largest red object–and it is also clear that the “indexical” difficulty which we found in the case of “object nearest to me” does not apply.
Nonetheless, there still seems to be something wrong with the idea that this sort of concept should form part of the definition of God. After all, if we had different responses, our concepts would be different, and then–on this picture–it seems that our definition could only be satisfied by a quite different entity. (Here, I imagine that eveything else is held constant–so that the same form of words is used to express evaluative concepts, but the content of those evaluations differ because the relevant responses differ.) Thus, I recognise a range of possible situations: there are cases in which our reponses are such as to lead the thesitic definition to pick out a unique entity, and there are cases in which only some other responses are such as to lead the tehistic definition to pick out a unique entity. Should I subscribe to the theistic definition? Well, what reason do I have for thinking that our use of the expression “morally perfect” is one of the privileged ones which does in fact lead the theistic definition to pick out a unique being?
I suppose that there are two sorts of answers which might be given at this point. First, one might try to argue on a posteriori grounds–i.e. to argue that available empirical evidence supports the view that the creator of the universe is morally perfect. This response seems to me to be on shaky ground. Doubtless, if God does exist, then it is reasonable to attribute some good actions, intentions, etc. to him. Moreover, this attribution seems to be perfectly compatible with a response-dependent analysis of moral concepts. But it also seems to me that the evidence adduced in traditional presentations of the problem of evil does not support the conclusion that God is morally perfect. That is, the available evidence strongly suggests that some sort of a priori justification of the attribution of moral perfection to God is required.
The second option which one might pursue is to try to argue on a priori grounds that God is morally perfect. But now, since it appears that there are no a priori arguments which turn on the actual content of our moral concepts, it seems that the same objection which we raised in the case of simple meta-ethical subjectivism takes hold. For example, if one were to argue that there is a priori good reason to suppose that the creator of the universe is perfect in every respect, one would not have provided any reason for thinking that God is morally perfect. A priori, it seems no less reasonable to suppose that some other range of responses would give rise to a concept of “moral perfection” which is truely applicable to God than it is to suppose that God is morally perfect. Surely we should be guilty of a very unchristian hubris to think that we are especially privileged in the pattern of responses to which we conform. After all, we are erring and fallible–and our patterns of moral response have undergone great changes across the millenia … .
I suppose that one answer which a theist might give is that God has so arranged things that our responses are of just the sort which is needed in order to ensure that our definition does pick him out. However, this answer seems to conflict with other theistic views about freedom of the will–and, in any case, would seem to be an entirely groundless claim. Surely we have no evidence at all to support the view that a certain range of evaluative responses are just the ones which we need to make in order to bring it about that the theistic definition of God is uniquely satisfied by an object in our universe.
In sum, then, the problem is that it seems that one could not really be justified in doing all of the following:
(i) adopting the traditional definition of God;
(ii) believing that God exists;
(iii) adopting a response-dependent account of moral concepts.
For the putative fact that moral concepts are response-dependent seems to leave us with no good reasons for thinking that there is an omniscient, ominpotent, eternal being who is also morally perfect. If, as I have suggested, questions of empirical evidence can be set aside, then it seems that the actual content of our moral concepts is irrelevant to arguments about the nature of God–and yet, on a response-dependent analysis, purely formal features of our moral concepts can give us no reason for thinking that God is morally perfect.
(There is a related and more venerable objection to the idea that God is morally perfect which might also be mentioned here. Numerous medieval writers held that it was impossible for finite and imperfect creatures like us to form an adequate conception of the infinite and perfect properties of God. It seems to me that there is something to this objection in the case of the attribution of the property of moral perfection to God given the recognition that ethical properties are response-dependent. For, given that questions of empirical evidence can be set aside, what reason can we have for supposing that God possesses a property which is defined in terms of our responses?)
Finally, in order to allay possible misgivings, I should point out that these considerations do nothing to undermine the plausibility of a response-dependent account of evaluative concepts. The possibility of other ranges of response need not do anything to undermine our confidence in the truth of our ordinary judgements about good and bad, right and wrong. However, it seems to me that the recognition of that possibility does serve to undermine any attempt to use those judgements as part of the foundation of a definition of “God”.
Given the above arguments, I think that it is reasonable to conclude that the traditional thesitic conception of God requires the assumption of meta-ethical realism. Any meta-ethical theory which gives a constitutive role to the subjective responses of human subjects will be vulnerable to the sorts of arguments which I have given. For any such theory produces a “relative” conception of goodness–i.e. a conception of goodness in which the content of that concept depends upon how we actually respond, what we actually do, etc. (Note, by the way, that the central issue is not about the right to speak of the “truth” of moral judgements. Both (quasi-realist) projectivist and response-dependent theories can licence talk of the truth of moral judgements. Rather, the central issue concerns the response-independence of moral judgements.)
Furthermore, I also think that it is reasonable to suppose that this conclusion does count against the traditional theistic conception of God. It is, I think, very hard to see how the motivating force of moral considerations can be captured in a meta-ethical theory in which the content of moral judgements is strongly independent of our actual responses, judgements and actions. However, as I said in the introduction, that is a topic for another occasion.
Finally, I should point out that the argument of this paper cannot also be used to support the view that those tradional theistic views are incompatible with fairly strong forms of meta-ethical objectivism. For there seems to be no barrier to the supposition that an a priori argument might be used to establish that God is necessarily morally perfect given that moral perfection is a response-independent property of the universe. The prospects for such arguments may be dim; but that is an entirely different issue.
“Is God Good By Definition?” is copyright © 1992 by Graham Oppy. All rights reserved.
The electronic version is copyright © 1998 by Internet Infidels with the written permission of Graham Oppy. All rights reserved.