Jan 5 JDN 2460681

There is another line of attack against a scientific approach to morality, one which threatens all the more because it comes from fellow scientists. Even though they generally agree that morality is real and important, many scientists have suggested that morality is completely inaccessible to science. There are a few different ways that this claim can be articulated; the most common are Stephen Jay Gould’s concept of “non-overlapping magisteria” (NOMA), David Hume’s “is-ought problem”, and G.E. Moore’s “naturalistic fallacy”. As I will show, none of these pose serious threats to a scientific understanding of morality.

NOMA

Stephen Jay Gould, though a scientist, an agnostic, and a morally upright person, did not think that morality could be justified in scientific or naturalistic terms. He seemed convinced that moral truth could only be understood through religion, and indeed seemed to use the words “religion” and “morality” almost interchangeably:

The magisterium of science covers the empirical realm: what the Universe is made of (fact) and why does it work in this way (theory). The magisterium of religion extends over questions of ultimate meaning and moral value. These two magisteria do not overlap, nor do they encompass all inquiry (consider, for example, the magisterium of art and the meaning of beauty).

If we take Gould to be using a very circumscribed definition of “science” to just mean the so-called “natural sciences” like physics and chemistry, then the claim is trivial. Of course we cannot resolve moral questions about stem cell research entirely in terms of quantum physics or even entirely in terms of cellular biology; no one ever supposed that we could. Yes, it’s obvious that we need to understand the way people think and the way they interact in social structures. But that’s precisely what the fields of psychology, sociology, economics, and political science are designed to do. It would be like saying that quantum physics cannot by itself explain the evolution of life on Earth. This is surely true, but it’s hardly relevant.

Conversely, if we define science broadly to include all rational and empirical methods: physics, chemistry, geology, biology, psychology, sociology, astronomy, logic, mathematics, philosophy, history, archaeology, anthropology, economics, political science, and so on, then Gould’s claim would mean that there is no rational reason for thinking that rape and genocide are immoral.

And even if we suppose there is something wrong with using science to study morality, the alternative Gould offers us—religion—is far worse. As I’ve already shown in previous posts, religion is a very poor source of moral understanding. If morality is defined by religious tradition, then it is arbitrary and capricious, and real moral truth disintegrates.

Fortunately, we have no reason to think so. The entire history of ethical philosophy speaks against such notions, and had Immanuel Kant and John Stuart Mill alive been alive to read them, they would have scoffed at Gould’s claims. I suspect Peter Singer and Thomas Pogge would scoff similarly today. Religion doesn’t offer any deep insights into morality, and reason often does; NOMA is simply wrong.

What’s the problem with “ought” and “is”?

The next common objection to a scientific approach to morality is the remark, after David Hume, that “one cannot derive an ought from an is”; due to a conflation with a loosely-related argument that G.E. Moore made later, the attempt to derive moral statements from empirical facts has become called the “naturalistic fallacy” (this is clearly not what Moore intended; I will address Moore’s actual point in a later post). But in truth, I do not really see where the fallacy is meant to lie; there is little difference in principle between deriving “ought” from “is” than there is from deriving anything from anything else.

First, let’s put aside direct inferences from “X is true” to “X ought to be true”; these are obviously fallacious. If that’s all Hume was saying, then he is of course correct; but this does little to undermine any serious scientific theory of morality. You can’t infer from “there are genocides” to “there ought to be genocides”; nor can you infer from “there ought to be happy people” to “there are happy people”; but nor would I or any other scientist seek to do so. This is a strawman of naturalistic morality.

It’s true that some people do attempt to draw similar inferences, usually stated in a slightly different form—but these are not moral scientists, they are invariably laypeople with little understanding of the subject. Arguments based on the claim that “homosexuality is unnatural” (therefore wrong) or “violence is natural” (therefore right) are guilty of this sort of fallacy, but I’ve never heard any credible philosopher or scientist support such arguments. (And by the way, homosexuality is nearly as common among animals as violence.)

A subtler way of reasoning from “is” to “ought” that is still problematic is the common practice of surveying people about their moral attitudes and experimentally testing their moral behaviors, sometimes called experimental philosophy. I do think this kind of research is useful and relevant, but it doesn’t get us as far as some people seem to think. Even if we were to prove that 100% of humans who have ever lived believe that cannibalism is wrong, it does not follow that cannibalism is in fact wrong. It is indeed evidence that there is something wrong with cannibalism—perhaps it is maladaptive to the point of being evolutionarily unstable, or it is so obviously wrong that even the most morally-blind individuals can detect its wrongness. But this extra step of explanation is necessary; it simply doesn’t follow from the fact that “everyone believes X is wrong” that in fact “X is wrong”. (Before 1900 just about everyone quite reasonably believed that the passage of time is the same everywhere regardless of location, speed or gravity; Einstein proved everyone wrong.) Moral realism demands that we admit people can be mistaken about their moral beliefs, just as they can be mistaken about other beliefs.

But these are not the only way to infer from “is” to “ought”, and there are many ways to make such inferences that are in fact perfectly valid. For instance, I know at least two ways to validly prove moral claims from nonmoral claims. The first is by conjunctive addition: “2+2=4, therefore 2+2=4 or genocide is wrong”. The second is by contradictory explosion: “2+2=5, therefore genocide is wrong”. Both of these arguments are logically valid. Obviously they are also quite trivial; “genocide is wrong” could be replaced by any other conceivable proposition (even a contradiction!), leaving an equally valid argument. Still, we have validly derived a moral statement from nonmoral statements, while obeying the laws of logic.

Moreover, it is clearly rational to infer a certain kind of “ought” from statements that entirely involve facts. For instance, it is rational to reason, “If you are cold, you ought to close the window”. This is an instrumental “ought” (it says what it is useful to do, given the goals that you have), not a moral “ought” (which would say what goals you should have in the first place). Hence, this is not really inferring moral claims from non-moral claims, since the “ought” isn’t really a moral “ought” at all; if the ends are immoral the means will be immoral too. (It would be equally rational in this instrumental sense to say, “If you want to destroy the world, you ought to get control of the nuclear launch codes”.) In fact this kind of instrumental rationality—doing what accomplishes our goals—actually gets us quite far in defining moral norms for real human beings; but clearly it does not get us far enough.

Finally, and most importantly, epistemic normativity, which any rational being must accept, is itself an inference from “is” to “ought”; it involves inferences like “Is it raining, therefore you ought to believe it is raining.”

With these considerations in mind, we must carefully rephrase Hume’s remark, to something like this:

One cannot nontrivially with logical certainty derive moral statements from entirely nonmoral statements.

This is indeed correct; but here the word “moral” carries no weight and could be replaced by almost anything. One cannot nontrivially with logical certainty derive physical statements from entirely nonphysical statements, nor nontrivially with logical certainty derive statements about fish from statements that are entirely not about fish. For all X, one cannot nontrivially with logical certainty derive statements about X from statements entirely unrelated to X. This is an extremely general truth. We could very well make it a logical axiom. In fact, if we do so, we pretty much get relevance logic, which takes the idea of “nontrivial” proofs to the extreme of actually considering trivial proofs invalid. Most logicians don’t go so far—they say that “2+2=5, therefore genocide is wrong” is technically a valid argument—but everyone agrees that such arguments are pointless and silly. In any case the word “moral” carries no weight here; it is no harder to derive an “ought” from an “is” than it is to derive a “fish” from a “molecule”.

Moreover, the claim that nonmoral propositions can never validly influence moral propositions is clearly false; the argument “Killing is wrong, shooting someone will kill them, therefore shooting someone is wrong” is entirely valid, and the moral proposition “shooting someone is wrong” is derived in large part from the nonmoral proposition “shooting someone will kill them”. In fact, the entire Frege-Geach argument against expressivism hinges upon the fact that we all realize that moral propositions function logically the same way as nonmoral propositions, and can interact with nonmoral propositions in all the usual ways. Even expressivists usually do not deny this; they simply try to come up with ways of rescuing expressivism despite this observation.

There are also ways of validly deriving moral propositions from entirely nonmoral propositions, in an approximate or probabilistic fashion. “Genocide causes a great deal of suffering and death, and almost everyone who has ever lived has agreed that suffering and death are bad and that genocide is wrong, therefore genocide is probably wrong” is a reasonably sound probabilistic argument that infers a moral conclusion based on entirely nonmoral premises, though it lacks the certainty of a logical proof.

We could furthermore take as axiom some definition of moral concepts in terms of nonmoral concepts, and then derive consequences of this definition with logical certainty. “A morally right action maximizes pleasure and minimizes pain. Genocide fails to maximize pleasure or minimize pain. Therefore genocide is not morally right.” Obviously one is free to challenge the definition, but that’s true of many different types of philosophical arguments, not a specific problem in arguments about morality.

So what exactly was Hume trying to say? I’m really not sure. Maybe he has in mind the sort of naive arguments that infer from “unnatural” to “wrong”; if so, he’s surely correct, but the argument does little to undermine any serious naturalistic theories of morality.