+Nevertheless, looking at how people behave in the world and trying to describe what you see is a form of _empiricism_, even if it's less third-party-legible than Science. In trying to upgrade our naïve empiricism to a Science, we should hope to design careful surveys to give us quantitative measurements of the qualitative patterns we see "in the wild", but getting this right is a surprisingly tricky endeavor. We're not obligated to throw away all our qualitative observations, in favor of single (!) survey question, just because the latter is quantifiable.
+
+You wrote about this in ["My IRB Nightmare"](https://slatestarcodex.com/2017/08/29/my-irb-nightmare/), expressing skepticism about a screening test for bipolar disorder:
+
+> You ask patients a bunch of things like "Do you ever feel really happy, then really sad?". If they say 'yes' to enough of these questions, you start to worry.
+
+> Some psychiatrists love this test. I hate it. Patients will say "Yes, that absolutely describes me!" and someone will diagnose them with bipolar disorder. Then if you ask what they meant, they'd say something like "Once my local football team made it to the Super Bowl and I was really happy, but then they lost and I was really sad." I don't even want to tell you how many people get diagnosed bipolar because of stuff like this.
+
+> There was a study that supposedly proved this test worked. But parts of it confused me, and it was done on a totally different population that didn't generalize to hospital inpatients.
+
+The reason it makes any sense at all for you to be skeptical, is because our beliefs about the existence and etiology of "bipolar disorder", don't completely stand or fall on this particular test. People _already_ had many observations pointing to the idea of "bipolar disorder" as a common cluster of symptoms. From your years of clinical experience, you think you know with your eyes what the cluster looks like. So when people whose favorite team lost the Super Bowl happen to answer "Yes" to the some of the same survey questions as people who you've _seen_ in the frenzy of mania and depths of depression, you generate the hypothesis: "Gee, maybe different populations are interpreting the question differently." Not as a _certainty_—maybe further research will provide more solid evidence that "bipolar disorder" isn't what you thought—but there's nothing un-Bayesian about thinking that your brain's pattern-matching capabilities are on to something important that this particular survey instrument isn't catching. You're not scientifically obligated to _immediately_ jump to "Bipolar Is Common and Not Especially Related to Mania or Depression."
+
+This shouldn't even be surprising when you consider the ambiguity and fuzziness of natural language: faced with a question, and prompted to give a forced-choice Yes/No or 1–5 response, people will assume the question was "meant for them" and try to map the words into some reference point in their experience. If the question _wasn't_ "meant for them"—if people who have never had a manic episode are given a set of questions formulated for a population of bipolar people—or if actual women are given a set of questions formulated for a population of males with a sex fantasy about being female—I think you _do_ get a lot of "Am I happy then sad sometimes? Sure, I guess so" out-of-distribution response behavior that doesn't capture what's actually going on.
+
+If you're wary that a survey about moods done on a totally different population might not generalize to hospital inpatients, I think you should be still more wary that that a survey _about sexuality_ might not generalize to people _of different sexes_! Even if you're skeptical of most evopsych accounts of psychological sex differences (there were no trucks or makeup in the environment of evolutionary adaptedness), sexuality is the _one domain_ where I think we have very strong prior reasons to expect cross-sex [empahtic inference](https://www.lesswrong.com/posts/9fpWoXpNv83BAHJdc/the-comedy-of-behaviorism) to fail.
+
+This is why I expect the standard "But cis women are autogynephilic too!!" cope to fall apart on further examination. I'm not denying the survey data itself (neither yours nor [Moser 2009](http://unremediatedgender.space/papers/moser-agp_in_women.pdf)'s); I'm saying we have enough prior knowledge about what females and males are like, to strongly suspect that women who answer Yes to the same survey questions as AGP males are mostly in the position of saying that they got really happy and then really sad when their team lost the Super Bowl.
+
+The reason this isn't special pleading that makes my theory unfalsifiable, is because my skepticism is _specifically_ about these mass survey questions where we haven't done the extra work to try to figure out whether the 1–5 question means the same thing to everyone; I'm happy to talk about qualitative predictions about what we see when we have a _higher-bandwidth_ channel into someone's mind than a 1–5 survey response.
+
+You _start_ to get into these issues with the "My literal body is arousing" account in the post, but I expect more digging to reveal a different picture. Of course, I can't know with any confidence (without more interviewing) what's really going on in [TODO: unredact name1]'s head. But to match one anecdote with another: when I _first_ came out to [TODO: name2] in 2016, her _first_ reaction was, "I can relate to this; I'm turned on by the idea of being a hot girl", but then when I went into _a little more detail_ (linking to the ["Man, I Feel Like a Woman"](https://tvtropes.org/pmwiki/pmwiki.php/Main/ManIFeelLikeAWoman) TVTropes page), she revised to, "Wait, no, I can't relate this at all; like, you _want_ to have breasts?" (I regret that this is paraphrased from memory; I haven't been able to find the transcript because I think Sarah deleted her Facebook account a couple times.)
+
+My common-sense intuition is that the experience of being happy and proud with one's own sexed body (which is pretty common and normal), and autogynephilic/autoandrophilic cross-sex fantasies (which are less common, but by no means rare), are superficially similar enough that they can generate overlapping reports if you _just_ ask "Would it be sexy to be her, 1–5", but that when you poke at the details, they're going to turn out to be _very_ different psychological phenomena that you shouldn't lump together as "autogenderphilia".
+
+Fundamentally, I just have _so much_ trouble _actually_ believing that [name1]'s experience of her body is more relevantly similar to mine than [name2]'s, _even if_ [name1] ends up sometimes using similar English words as me (_e.g._, "it's hot that I have breasts"). I can conceive of being wrong about this, but I don't think the _SSC_ survey data is a powerful enough instrument to make that call—I'd want in-depth interviews and preferably the kind of physical arousal measurements that Michael Bailey's lab does.
+
+In a world where it was _actually true_ where "if you identify as a gender, and you're attracted to that gender [...]" reflected the causal structure of what was actually going on in the world, I would expect the things trans lesbians say to each other in naturalistic contexts when the general public isn't looking, to _look like_ the things cis lesbians say to each other in naturalistic contexts—and that's not what I see.
+
+Here's [an example from Twitter](https://web.archive.org/web/20210903211904/https://twitter.com/lae_laeta/status/1433880523160567808)—
+
+> The eternal trans lesbian question: So do I want to be her, or do I want to be with her?
+
+> The answer: Yes
+
+I see this "want her or want to be her" sentiment from trans women _and_ non-transitioned AGP men _very_ frequently. I can't recall any instances of cis lesbians saying this. The poster herself seems to implicitly acknowledge this, by calling it a "trans lesbian question" rather than merely a "lesbian" question!
+
+I think the boring hypothesis here is "Yes, of course, because trans women are AGP men, which are not the same thing as actual lesbians." Again, this isn't Science, because I'm just using my brain's pattern-matching capabilities (I could be selectively remembering, distorting my categories, _&c._). With time and funding, I'm sure it would be possible to make it more formal—gather Reddit comments from cis and trans women, have raters categorize themes while blinded to the cis/trans identity of the authors ...
+
+But I begin to despair this is a domain where [Science can't help](https://www.lesswrong.com/posts/wzxneh7wxkdNYNbtB/when-science-can-t-help). It seems like people mostly _agree_ about empirical observations! From my perspective, it looks like the _Slate Star_/Alicorner crowd manages come up with these absurdly gerrymandered verbal explanations that can't _possibly_ match up with the machinery your brain must be using to know what to anticipate, but if you don't see this after it's already been pointed out, then I'm not sure how to proceed. From my perspective, it looks like you just have a fundamentally broken epistemology; from your perspective, I probably look like I'm dogmatically making unexplained inferential leaps.
+
+Ozy has an old post about [how "the community" doesn't have a _gender_ gap; we merely have an _assigned sex at birth_ gap](https://thingofthings.wordpress.com/2014/12/01/lw-has-an-assigned-sex-at-birth-gap-not-a-gender-gap/). In my worldview, this should be _embarrassing_. (If you keep running into domains where "assigned" sex is a more useful predictor than "gender", that should be a clue that sex is real and gender identity is fake.) But if Ozy's mind hasn't been [created already in motion](https://www.lesswrong.com/posts/CuSTqHgeK4CMpWYTe/created-already-in-motion) to find it embarrassing even after it's been pointed out, then I'm not sure what else I can say?
+
+And yet—if it were _just_ a matter of different priors (where my stronger [inductive bias](https://en.wikipedia.org/wiki/Inductive_bias) lets me learn faster from less data, at the cost of [being wrong in universes that I think mostly don't exist](https://en.wikipedia.org/wiki/No_free_lunch_in_search_and_optimization)), I would expect you
+
+
+
+
+
+
+
+
+
+
+I would totally respect it if you were merely _uncertain_ about the AGP→gender-ID _vs._ gender-ID→AGP causality; I can't expect everyone to share my parsimony intuitions.
+
+But on Discord, you said "it just seemed totally wrong"!