Yes, You Can Argue About Definitions

Contra Scott Alexander a little bit

One of my favorite Scott Alexander essays is Social Justice And Words, Words, Words. It’s mostly about irritating and obfuscatory word games played by those of social justice persuasion but has a small section that strikes me as clearly wrong:

There’s no disputing the definitions of words. If you say that “racism” is a rare species of nocturnal bird native to New Guinea which feeds upon morning dew and the dreams of young children, then all I can do is point out that the dictionary and common usage both disagree with you. And the sources I cited above have already admitted that “the dictionary is wrong” and “no one uses the word racism correctly”.

Now, the article spends a lot of time doing stuff that for all the world looks like arguing about words. Pointing out that the dictionary and common usage disagree with someone is a way of arguing about definitions. But seeing as this idea that you can’t argue about definitions seems to be pretty widely-believed—including by people as clever and insightful as Scott—I thought I would explain why this view is wrong.

When we argue about definitions, we can argue about one of two things. First, how some words are actually used. Second, how some words should be used. Both of these are things that can be argued about.

Regarding the first category, people can argue about how words are actually used by adducing intuitive counterexamples to definitions. For a while, most philosophers thought that knowledge was justified true belief. But then Gettier gave a counterexample to this: if you have two cents in your pocket, and are justified in thinking you’ll be hired for some job, you can justifiably think “the person who will be hired for this job will have two cents in his pocket.” Now, suppose it turns out that someone else gets hired—but he had two cents in his pocket. Your belief that the person who will be hired would have two cents in his pocket was true and it was justified—but it seems like it wasn’t knowledge. So knowledge isn’t justified true belief.

This sure looks like arguing about what words mean. One person proposes a definition, someone else gives a counterexample, and they argue about which is overall more convincing. Such arguments are quite common in philosophy. If this is not arguing about definitions, I don’t know what is!

The second way of arguing about definitions is arguing about how words should be used. For example, suppose we found out that every time a person uses the word “table” in the standard way, to describe the piece of furniture with legs and a flat surface, 100,000 cute puppies in distant galaxies were thrown into a furnace. One could quite sensibly, in such a world, have arguments for why people should stop using the word table as they had been.

For a more realistic case, there’s lots of debate about how the words “man” and “woman” should be used. Some people argue that this should include trans women, on grounds that this has desirable social effects. Others disagree. Whatever one thinks about this subject, it’s clearly the kind of thing that people can argue about! People do, in fact, routinely argue about it—so clearly they can do it! People rarely do things that are literally impossible.

How does this work? Exactly the same way any other arguments about morality work. People can, in fact, argue about morality—and frequently do! I don’t think morality is subjective, but even if it is, it’s still the kind of subjective thing that can be argued about.

Now, it’s perfectly all-right to have a stipulative definition. If I say “for the purposes of this paper, I will use effective altruism to mean such and such,” that’s perfect alright (take notes Leif Wenar). It does not require any broader commitments to either that being common usage or to that being an ideal way of defining the term. There the author is just talking about how they personally are using it. It’s a bit like when someone says in a paper “call this principle Non-extreme Anti-Repugnance,” or something—they’re just giving a label to some particular idea.

But it does not follow from the fact that words can be used stipulatively, in a way that makes arguments unproductive, that you can’t argue about words. There are lots of contexts in ordinary life when you can argue about words and it is indeed productive to do so. When one argues about whether trans women are women, they are arguing about the use of words. And there can be arguments about whether such a view is correct!

You can also use a moral theory stipulatively. You can say, “for the purpose of this paper, I will assume utilitarianism.” You can also do this with factual statements—saying, for instance, “for the purpose of this essay, I will assume atheism, and the many-worlds interpretation.” This doesn’t mean you can’t argue about values and facts. It just means there’s a context in which a person isn’t really taking a view on what values or facts mean and so it’s inappropriate to argue with them.

So if someone is using definitions stipulatively, fighting over the definitions is unproductive. But if not, criticizing them is fair game. You can actually criticize the person who claims that racism is a species of nocturnal New Guinean bird—and then demands we eradicate on grounds of anti-racism, before saying that if you criticize their bird eradication you hate black people.