Trans Women Are Women
But it's cool if you disagree (even though you shouldn't)
1. Are Trans Women Women?
recently debated a man named Chris Date who is, above all, a follower of Jesus Christ, about whether trans women were women, and whether trans men were men.He answered that they were—that “man” and “woman” refer to something other than strictly biological sex. Specifically, he thought they referred to some more nebulous sociocultural concept incorporating secondary sex characteristics, interests, tendencies, behaviors, and so on—broadly speaking, “gender.” So, he argued, insofar as trans women fulfill the sociocultural roles we generally expect women to fill, they are, in fact, women.
Date took the other side: he made lots of points about historical and legal precedents which explicitly equated the terms “man” and “woman” with “male” and “female.”1 Interestingly, at one point he appealed to the field of linguistics—he mentioned that “man” and “woman” both appear on the Swadesh list, a collection of 100 concepts common across linguistic groups, where they’re defined as “adult human male” and “adult human female” respectively. This, he argued, was fairly conclusive evidence that trans women couldn’t be women—the most fundamental and universal meaning of the word “woman” excludes them.
Bentham pushed back, comparing the words “man” and “woman” to the word “parent.” He argued that while it initially referred to biological parentage only, we’ve since revised it to include step-parents and adoptive parents who serve many of the same social roles as biological parents. It’d be silly to tell a loving step-father he can’t pick his kid up from school because, technically, “parent” comes from the Latin “pariō” meaning “to bring forth,” and he never actually brought that kid forth from his loins, so, sorry, tell mom to come in next time.
“Parent” is sociocultural in meaning, and so are “man” and “woman.”
Date countered: “parent” doesn’t appear on the Swadesh list! Fine, it can be sociocultural, but it’s not universal in the same way that “man” and “woman” are.
The two continued to debate a million other points, but a commenter on Bentham’s blog pointed out after the fact: right after “man” and “woman” on the now-more-widely-accepted Swadesh 207-word list appear “child,” “wife,” “husband,” “mother,” and “father.” Clearly, sociocultural constructs can be fundamental and universal too!
I think this points at something pretty important, and generally underrated in debates like these: definitions are exactly as good as they are useful. And I don’t mean this in a normative sense—we’ll get to that soon, Mr.
—but in a pure and mathematical one. A definition is worth its weight in bits.Sociocultural definitions can often be extremely useful: if a step-father looks, swims, and quacks like a father, it’s clearest if you just call him one. When you say something like, “that man is my father,” I assume a number of things about your relationship with him: that he raised you, that he cares about you, that he has or had some relationship with your mother, etc. Maybe I’ll also assume he’s your biological father—and maybe that assumption will be wrong! But, ultimately, in the way that I’ll subsequently act toward the two of you, that assumption will matter a lot less than all the other things. Even if my belief that the two of you are biologically related is false, I should let him sign you out of school and drive you home.
Ah, you say. But what if you’re a doctor asking for a kidney donation or something? Then it would be very bad if you mistakenly assumed the man was my biological father!
This is true! In certain contexts, calling a step-father a father isn’t very helpful, isn’t very information-rich. If I’m a doctor asking you who your father is, presumably I’m referring to the concept GUY WHO BIRTHED YOU rather than GUY WHO’S RESPONSIBLE FOR DRIVING YOU AROUND AND FEEDING YOU, and so the best reply will point to the former.
The parallel should, at this point, be clear: in most contexts, when I ask, “is that person a woman?” about a trans woman, I’m asking whether or not she fits into VAGUE SOCIOCULTURAL CLUSTER OF WOMANY THINGS, not whether she fits into GROUP OF PEOPLE WHO PRODUCE LARGE GAMETES, and so the most helpful thing for you to say is “yes.” (Which means it’s the correct thing for you to say!)
Maybe in certain contexts—if I’m a director of a women’s swimming team, for example—I actually am referring more to the sexual category, and your answer should change. “That person is a woman for most intents and purposes, but probably shouldn’t be competing alongside biological women in a sports league” is a perfectly coherent stance.
Cool, everyone agreed? Political discourse solved and over now?
2. What if I’m Not a Mindreader?
Wow, well, sucks to be you I guess.
This is a good objection, though. It’s not always immediately obvious what the most helpful definition to provide is when someone asks you a question like “what is a woman?”—maybe you don’t know who’s asking, maybe they’re being intentionally vague, or maybe they’re trying to provoke you on the internet and won’t accept any definition but the one they agree with. In a situation like this, what should we do?
Well, since I’m defining “right” as “most useful” and “most useful” in a fairly strict information theory–heavy sense, we can just define a mathematical function to answer this question.
In fact, a team of natural language researchers at Google took this project on in 2013, and they did a pretty spanking job. The researchers collected a massive English-language corpus and encoded it into vector space—they were able to cluster semantically-similar words like “university” and “college” physically (ok, mathematically) right next to one another. Word2vec, as they called the project, even preserved semantic relationships between words. So you could, for example, take a word like “woman,” subtract the values for “adult,” “human,” and “female,” and see what was left. Then maybe add context, “medical” or “sports” or “bathroom” and see what cluster you landed in.2
Some other researchers figured out that you could use these relationships to do analogies too. Take the value for the word “France,” subtract “Paris,” then add “Tokyo,” and you would get “Japan.” Cool! You could even do “king” - “man” + “woman,” and Word2vec would spit out “queen.”
And then someone realized you could also do “doctor” - “man” + “woman” and you’d get “nurse.” And they found out that “man” - “computer programmer” + “woman” = “homemaker.” And so on…
Sexism was baked deeply into the English-language canon! What Word2vec found was that “doctor” wasn’t only transmitting information like “sees patients” and “practices medicine,” but also “is probably a dude.” Not so great!
The method I’ve described for defining words is completely descriptive—it accepts use as it is, even if that use is (intentionally or not) harmful, discriminatory, and dumb. Can we do better? Should we be a bit more prescriptive about language use?
3. Silas Gets Woke as Fuck
A day or two before Bentham’s debate, friend of the blog Silas Abrahamsen had bravely sparred with the dastardly
over philosophy of language and what it means to be a woman. So he was able to quickly fire off his total-semantic-skepticist take on what it means to be a woman:
I agree with Silas that words don’t have singular “Real Meanings”—pretty much every one we use is a rough gesture at lots of different ideas. But I still think some meanings can be more correct than others—Silas imagines using the word “oatmilk” to refer to the concept of WATER, and claims that this would be fundamentally unproblematic. He’d just be a weirdo, not a wrong weirdo, because there’s no Really Actually True Truth to be wrong about.
And then he also writes that this doesn’t actually have any impact on the usefulness of conversation:
So communication works like this: I hope to convey some thoughts to you. To do this, I should use some words, and through experience I have good reason to think that using such and such words will cause you to have roughly the thoughts. Thus I choose to use these words. On your end, you know that such and such words are usually meant to convey these thoughts, and so you interpret them as such. Why should I also add that it is Actually True™ that these words mean this? There’s just no need!
I… agree? I think?
It seems like he and I think broadly the same thing: some ways of using words to relate to concepts are more practical than others, and yet there’s no way to say that any given use is Ultimately Unforgivably Wrong. So, cool!
Here’s where Silas starts to go off the deep end, though:
You shouldn’t go around hurting people’s feelings for no good reasons, and seeing as you’re not making a mistake by including trans women, there is no good reason. Actually, even if you were making a factual mistake, you should probably still include trans women, all things considered.
Au contraire, mon frère! (Sorry if that’s hard for you to understand: Silas is Danish, so I’m trying to be #Inclusive.)
While I agree that needless feeling-hurting is dumb, we obviously can’t just go around calling anyone whatever they want to be called! For instance, if you identify as an attack helicopter, even if it’ll hurt your feelings a bit for society to deny redefining “attack helicopter” to include you, it probably shouldn’t do that given that infecting every use of “attack helicopter” with a bit of the PERSONAL IDENTITY concept is going to mean a lot of bit-loss.
(I think similar arguments apply to
’s L take on algospeak: a word like “murder” is generally informative, whereas “unalive” is generally confusing, and confusion is the exact opposite of the point of communication! Also it’s so dumb-sounding, I really hate it, Goddamn. Oughta be enough.)In the case of trans women, Silas’ argument probably goes through—but only because the least-lossy definition of “woman” almost certainly includes Blair White:
And excludes Buck Angel:
Like, duh.
4. Don’t Be Red-Pilling on Trans Acceptance Though
But Silas seemingly wants to go further—he’s even written (admittedly ages ago in 2024) that we should probably be doing lots of conceptual engineering all the time. I think this is a bad and wrong lesson to walk away with.
There’s probably a case to be made that the single most defining feature of the woke moment was its fixation on conceptual engineering.
“No, ‘racist’ doesn’t actually refer to PREJUDICED PERSON, you ignoramus, it refers to ANYONE WHO BENEFITS AT ALL FROM THE SYSTEMS OF PERSECUTION THAT HAVE BEEN RAMPANT IN AMERICAN SOCIETY SINCE 1619 AND ISN’T ACTIVELY WORKING TO TEAR OUR INSTITUTIONS TO THE GROUND. And if it doesn’t totally mean that yet, then we should be working very hard to make it so! We need to put that definition in the dictionary, drill it into everyone’s mind, and cancel anyone who resists. ‘Cause it’s not strictly wrong—and really it’d clearly be much better if everyone was paying all their attention to systemic racism.”
Beyond my normal anti-radicalist points about humility, I should emphasize that there’s a really big glaring problem with this strategy: it doesn’t work at all!
The more you admonish the majority for wrongthink, the more that majority views you and your cause as the enemy, the more it realizes its supposed wrongthink is the norm, the more it gets mad at out-of touch elites, the more it elects Trump, and the more you end up with crazy authoritarian atrocities and terribleness.
Beating people over the head with “trans women are women because only bad, unethical people think otherwise” is absolutely the worst possible argument to make! I agree, kindness is nice! But if you want people to be kind, frankly, you can just say that—say, “it’d be nice not to misgender people”—rather than making the much stronger and much more invasive ask that doubters actually believe in transgender self-id.
Gay marriage eventually became legal not because Middle America was browbeaten over opposing it, but because Middle Americans met gay people, got to know them, realized they were just normal good people, and decided to be nice to them. The winning playbook is humanization and decency, not conceptual engineering and moralizing admonishment.
I believe that trans women are women. I think you should too.
But it’s also ok if you don’t—all you need to believe is that trans women are people. They’re generally good and decent and it costs you very little to be good and decent back, whether you deeply believe them to be women or not.
Bentham’s arguments against historical and legal precedents were convincing, I’d summarize as follows:
A) For a long time, “woman” has been totally underdetermined. 99.9% of the time, Bentham’s and Date’s definitions will agree—only recently has a gap opened up between them. Even if Bentham were “right” about what it means to be a woman, we should expect past sources to have mixed the word up with “adult human female” all the time.
B) Legal definitions are a totally different ballgame, and pretty unrelated to the task of conceptual analysis. There’s probably plenty of room for defining “woman” as “adult human female” in some contexts (or simply using the words “adult human female”) and defining it as a sociocultural construct in others.
I don’t know of anyone actually doing this, and would be interested to see what happened if they did. If you’re as techy or more than me, let me know what it might take to do this… (or if you want to do it for me).
I agree with the overall thrust of the approach, and I think Scott Alexander articulated it really nicely way back in 2014, before "what is a woman?" was quite the hot button question it became:
https://slatestarcodex.com/2014/11/21/the-categories-were-made-for-man-not-man-for-the-categories/
That said, I probably disagree with your claim that in most contexts, if I ask "is that person a woman?" to someone who knows that the person in question is a trans woman, the most helpful answer for them to give is "yes". I also don't think the most helpful answer is "no". Rather, I think the most helpful answer is "she is a trans woman."
I think a natural correlate of the approach is that lots of binary classifiers that are reasonably useful in many contexts are less useful when applied to edge cases, and the sensible thing to do when faced with those edge cases is not to insist on deciding whether the binary classifier applies or not, but just to switch to a more fine-grained classification system. This is a pretty standard thought about vague language; if you're asked whether someone is tall, many times the most helpful answer is "yes" or "no", but for people on the border between being tall and not, it's often better to just say roughly how tall they are rather than trying to decide whether that height is enough to count as tall.
Same goes with asking whether some country is a democracy; for many countries, the helpful answer is "yes" or "no", but sometimes it's vague enough that it's better to refuse to answer the yes/no question, and say something about what the political system is like in a way that gives the listener an idea of why the yes/no question isn't straightforward.
There are a variety of interests people might have in inquiring whether someone is a woman, some of which would lead to grouping trans women with women, and some of which would not. Given that we have this other category, "trans woman", it seems to me almost always better to switch to that than to try to answer the yes/no question.
"Gay marriage eventually became legal not because Middle America was browbeaten over opposing it, but because Middle Americans met gay people, got to know them, realized they were just normal good people, and decided to be nice to them."
I disagree. It was not due to meeting gay people. It was due to gay people being featured positively in media, especially in the music industry -- Freddie Mercury and Elton John stand out. Madonna promoted gay men heavily. Gay people have been "around" to "meet" before their image improved. It was top-down engineering from Hollywood, not spontaneous encounters on the street. If you don't actively promote LGBTQ, people aren't going to naturally accept it due to cultural inertia and mass conservatism.