Jan 12 JDN 2460688

In last week’s post I talked about some of the arguments against ethical naturalism, which have sometimes been called “the naturalistic fallacy”.

The “naturalistic fallacy” that G.E. Moore actually wrote about is somewhat subtler; it says that there is something philosophically suspect about defining something non-natural in terms of natural things—and furthermore, it says that “good” is not a natural thing and so cannot be defined in terms of natural things. For Moore, “good” is not something that can be defined with recourse to facts about psychology, biology or mathematics; “good” is simply an indefinable atomic concept that exists independent of all other concepts. As such Moore was criticizing moral theories like utilitarianism and hedonism that seek to define “good” in terms of “pleasure” or “lack of pain”; for Moore, good cannot have a definition in terms of anything except itself.

My greatest problem with this position is less philosophical than linguistic; how does one go about learning a concept that is so atomic and indefinable? When I was a child, I acquired an understanding of the word “good” that has since expanded as I grew in knowledge and maturity. I need not have called it “good”: had I been raised in Madrid, I would have called it bueno; in Beijing, hao; in Kyoto, ii; in Cairo, jaiid; and so on.

I’m not even sure if all these words really mean exactly the same thing, since each word comes with its own cultural and linguistic connotations. A vast range of possible sounds could be used to express this concept and related concepts—and somehow I had to learn which sounds were meant to symbolize which concepts, and what relations were meant to hold between them. This learning process was highly automatic, and occurred when I was very young, so I do not have great insight into its specifics; but nonetheless it seems clear to me that in some sense I learned to define “good” in terms of things that I could perceive. No doubt this definition was tentative, and changed with time and experience; indeed, I think all definitions are like this. Perhaps my knowledge of other concepts, like “pleasure”, “happiness”, “hope” and “justice”, is interconnected with “good” in such a way that none can be defined separately from the others—indeed perhaps language itself is best considered a network of mutually-reinforcing concepts, each with some independent justification and some connection to other concepts, not a straightforward derivation from more basic atomic notions. If you wish, call me a “foundherentist” in the tradition of Susan Haack; I certainly do think that all beliefs have some degree of independent justification by direct evidence and some degree of mutual justification by coherence. Haack uses the metaphor of a crossword puzzle, but I prefer Alison Gopnik’s mathematical model of a Bayes net. In any case, I had to learn about “good” somehow. Even if I had some innate atomic concept of good, we are left to explain two things: First, how I managed to associate that innate atomic concept with my sense experiences, and second, how that innate atomic concept got in my brain in the first place. If it was genetic, it must have evolved; but it could only have evolved by phenotypic interaction with the external environment—that is, with natural things. We are natural beings, made of natural material, evolved by natural selection. If there is a concept of “good” encoded into my brain either by learning or instinct or whatever combination, it had to get there by some natural mechanism.

The classic argument Moore used to support this position is now called the Open Question Argument; it says, essentially, that we could take any natural property that would be proposed as the definition of “good” and call it X, and we could ask: “Sure, that’s X, but is it good?” The idea is that since we can ask this question and it seems to make sense, then X cannot be the definition of “good”. If someone asked, “I know he is an unmarried man, but is he a bachelor?” or “I know that has three sides, but is it a triangle?” we would think that they didn’t understand what they were talking about; but Moore argues that for any natural property, “I know that is X, but is it good?” is still a meaningful question. Moore uses two particular examples, X = “pleasant” and X = “what we desire to desire”; and indeed those fit what he is saying. But are these really very good examples?

One subtle point that many philosophers make about this argument is that science can discover identities between things and properties that are not immediately apparent. We now know that water is H₂O, but until the 19th century we did not know this. So we could perfectly well imagine someone asking, “I know that’s H₂O, but is it water?” even though in fact water is H₂O and we know this. I think this sort of argument would work for some very complicated moral claims, like the claim that constitutional democracy is good; I can imagine someone who was quite ignorant of international affairs asking: “I know that it’s constitutional democracy, but is that good?” and be making sense. This is because the goodness of constitutional democracy isn’t something conceptually necessary, it is an empirical result based on the fact that constitutional democracies are more peaceful, fair, egalitarian, and prosperous than other governmental systems. In fact, it may even be only true relative to other systems we know of; perhaps there is an as-yet-unimagined governmental system that is better still. No one thinks that constitutional democracy is a definition of moral goodness. And indeed, I think few would argue that H₂O is the definition of water; instead the definition of water is something like “that wet stuff we need to drink to survive” and it just so happens that this turns out to be H₂O. If someone asked “is that wet stuff we need to drink to survive really water?” he would rightly be thought talking nonsense; that’s just what water means.

But if instead of the silly examples Moore uses, we take a serious proposal that real moral philosophers have suggested, it’s not nearly so obvious that the question is open. From Kant: “Yes, that is our duty as rational beings, but is it good?” From Mill: “Yes, that increases the amount of happiness and decreases the amount of suffering in the world, but is it good?” From Aristotle: “Yes, that is kind, just, and fair, but is it good?” These do sound dangerously close to talking nonsense! If someone asked these questions, I would immediately expect an explanation of what they were getting at. And if no such explanation was forthcoming, I would, in fact, be led to conclude that they literally don’t understand what they’re talking about.

I can imagine making sense of “I know that has three sides, but is it a triangle?”in some bizarre curved multi-dimensional geometry. Even “I know he is an unmarried man, but is he a bachelor?” makes sense if you are talking about a celibate priest. Very rarely do perfect synonyms exist in natural languages, and even when they do they are often unstable due to the effects of connotations. None of this changes the fact that bachelors are unmarried men, triangles have three sides, and yes, goodness involves fulfilling rational duties, alleviating suffering, and being kind and just (Deontology, consequentialism, and virtue theory are often thought to be distinct and incompatible; I’m convinced they amount to the same thing, which I’ll say more about in later posts.).

This line of reasoning has led some philosophers (notably Willard Quine) to deny the existence of analytic truths altogether; on Quine’s view even “2+2=4” isn’t something we can deduce directly from the meaning of the symbols. This is clearly much too strong; no empirical observation could ever lead us to deny 2+2=4. In fact, I am convinced that all mathematical truths are ultimately reducible to tautologies; even “the Fourier transform of a Gaussian is Gaussian” is ultimately a way of saying in compact jargon some very complicated statement that amounts to A=A. This is not to deny that mathematics is useful; of course mathematics is tremendously useful, because this sort of compact symbolic jargon allows us to make innumerable inferences about the world and at the same time guarantee that these inferences are correct. Whenever you see a Gaussian and you need its Fourier transform (I know, it happens a lot, right?), you can immediately know that the result will be a Gaussian; you don’t have to go through the whole derivation yourself. We are wrong to think that “ultimately reducible to a tautology” is the same as “worthless and trivial”; on the contrary, to realize that mathematics is reducible to tautology is to say that mathematics is undeniable, literally impossible to coherently deny. At least the way I use the words, the statement “Happiness is good and suffering is bad” is pretty close to that same sort of claim; if you don’t agree with it, I sense that you honestly don’t understand what I mean.

In any case, I see no more fundamental difficulty in defining “good” than I do in defining any concept, like “man”, “tree”, “multiplication”, “green” or “refrigerator”; and nor do I see any point in arguing about the semantics of definition as an approach to understanding moral truth. It seems to me that Moore has confused the map with the territory, and later authors have confused him with Hume, to all of our detriment.