-But, I argued, political convenience came at a dire cost to [our common interest](https://www.lesswrong.com/posts/4PPE6D635iBcGPGRy/rationality-common-interest-of-many-causes). There was a proverb Yudkowsky [had once failed to Google](https://www.lesswrong.com/posts/K2c3dkKErsqFd28Dh/prices-or-bindings), which ran something like, "Once someone is known to be a liar, you might as well listen to the whistling of the wind."
-
-Similarly, once someone is known to [vary](https://slatestarcodex.com/2014/08/14/beware-isolated-demands-for-rigor/) the epistemic standards of their public statements for political convenience—if they say categorizations can be lies when that happens to help their friends, but seemingly deny the possibility of categorizations being lies when that happens to make them look good politically ...
-
-Well, you're still _somewhat_ better off listening to them than the whistling of the wind, because the wind in various possible worlds is presumably uncorrelated with most of the things you want to know about, whereas [clever arguers](https://www.lesswrong.com/posts/kJiPnaQPiy4p9Eqki/what-evidence-filtered-evidence) who [don't tell explicit lies](https://www.lesswrong.com/posts/xdwbX9pFEr7Pomaxv/) are constrained in how much they can mislead you. But it seems plausible that you might as well listen to any other arbitrary smart person with a bluecheck and 20K followers. I know you're very busy; I know your work's important—but it might be a useful exercise, for Yudkowsky to think of what he would _actually say_ if someone with social power _actually did this to him_ when he was trying to use language to reason about Something he had to Protect?
-
-(Note, my claim here is _not_ that "Pronouns aren't lies" and "Scott Alexander is not a racist" are similarly misinformative. Rather, I'm saying that whether "You're not standing in defense of truth if you insist on a word, brought explicitly into question, being used with some particular meaning" makes sense _as a response to_ "X isn't a Y" shouldn't depend on the specific values of X and Y. Yudkowsky's behavior the other month had made it look like he thought that "You're not standing in defense of truth if ..." _was_ a valid response when, say, X = "Caitlyn Jenner" and Y = "woman." I was saying that, whether or not it's a valid response, we should, as a matter of [local validity](https://www.lesswrong.com/posts/WQFioaudEH8R7fyhm/local-validity-as-a-key-to-sanity-and-civilization), apply the _same_ standard when X = "Scott Alexander" and Y = "racist.")
-
-Without disclosing any specific content from private conversations with Yudkowsky that may or may not have happened, I think I _am_ allowed to say that our posse did not get the kind of engagement from Yudkowsky that we were hoping for. (That is, I'm Glomarizing over whether Yudkowsky just didn't reply, or whether he did reply and our posse was not satisfied with the response.)
-
-Michael said that it seemed important that, if we thought Yudkowsky wasn't interested, we should have common knowledge among ourselves that we consider him to be choosing to be a cult leader.
-
-Meanwhile, my email thread with Scott got started back up again, although I wasn't expecting anything to come out of it. I expressed some regret that all the times I had emailed him over the past couple years had been when I was upset about something (like psych hospitals, or—something else) and wanted something from him, which was bad, because it was treating him as a means rather than an end—and then, despite that regret, continued prosecuting the argument (Subject: "twelve short stories about language").
-
-One of Alexander's [most popular _Less Wrong_ posts ever had been about the noncentral fallacy, which Alexander called "the worst argument in the world"](https://www.lesswrong.com/posts/yCWPkLi8wJvewPbEp/the-noncentral-fallacy-the-worst-argument-in-the-world): for example, those who crow that abortion is _murder_ (because murder is the killing of a human being), or that Martin Luther King, Jr. was a _criminal_ (because he defied the segregation laws of the South), are engaging in a dishonest rhetorical maneuver in which they're trying to trick their audience into attributing attributes of the typical "murder" or "criminal" onto what are very noncentral members of those categories.
-
-_Even if_ you're opposed to abortion, or have negative views about the historical legacy of Dr. King, this isn't the right way to argue. If you call Janie a _murderer_, that causes me to form a whole bunch of implicit probabilistic expectations—about Janie's moral character, about the suffering of victim whose hopes and dreams were cut short, about Janie's relationship with the law, _&c._—most of which get violated when you subsequently reveal that the murder victim was a four-week-old fetus.
-
-Thus, we see that Alexander's own "The Worst Argument in the World" is really complaining about the _same_ category-gerrymandering move that his "... Not Man for the Categories" comes out in favor of. We would not let someone get away with declaring, "I ought to accept an unexpected abortion or two deep inside the conceptual boundaries of what would normally not be considered murder if it'll save someone's life." Maybe abortion _is_ wrong and relevantly similar to the central sense of "murder", but you need to make that case _on the empirical merits_, not by linguistic fiat.
-
-... Scott still didn't get it. He said that he didn't see why he shouldn't accept one unit of categorizational awkwardness in exchange for sufficiently large utilitarian benefits. He made an analogy to some [Glowfic](https://www.glowfic.com/) lore, a story about orcs who had unwisely sworn a oath to serve the evil god Melkor. Though the orcs intend no harm of their own will, they're magically bound to obey Melkor's commands and serve as his terrible army or else suffer unbearable pain. Our heroine comes up with a solution: she founds a new religion featuring a deist God who also happens to be named Melkor. She convinces the orcs that since the oath didn't specify _which_ Melkor, they're free to follow her new God instead of evil-Melkor, and the magic making the oath binding apparently accepts this casuistry if the orc themelf does.
-
-Scott's attitude towards the new interpretation of the oath in the story was analogous to his thinking about transgenderedness: sure, the new definition may be a little awkward and unnatural in some sense, but it's not literally objectively false, and it made life better for so many orcs. If [rationalists should win](https://www.lesswrong.com/posts/6ddcsdA2c2XpNpE5x/newcomb-s-problem-and-regret-of-rationality), then the true rationalist in this situation is the one who thought up this clever hack to save an entire species.
-
-I started drafting a long reply—but then I remembered that in recent discussion with my posse about what we might have done wrong in our attempted outreach to Yudkowsky, the idea had come up that in-person meetings are better for updateful disagreement-resolution. Would Scott be up for meeting in person some weekend? Non-urgent. Ben would be willing to moderate, unless Scott wanted to suggest someone else, or no moderator.
-
-... Scott didn't want to meet. At this point, I considered resorting to the tool of cheerful prices again, which I hadn't yet used against Scott—to say, "That's totally understandable! Would a financial incentive change your decision? For a two-hour meeting, I'd be happy to pay up to $4000 to you or your preferred charity. If you don't want the money, then sure, yes, let's table this. I hope you're having a good day." But that seemed sufficiently psychologically coercive and socially weird that I wasn't sure I wanted to go there. On 18 March, I emailed my posse asking what they thought—and then added that maybe they shouldn't reply until Friday, because it was Monday, and I really needed to focus on my dayjob that week.
-
-This is the part where I began to ... overheat. I tried ("tried") to focus on my dayjob, but I was just _so angry_. Did Scott _really_ not understand the rationality-relevant distinction between "value-dependent categories as a result of caring about predicting different variables" (as explained by the _dagim_/water-dwellers _vs._ fish example) and "value-dependent categories _in order to not make my friends sad_"? I thought I was pretty explicit about this? Was Scott _really_ that dumb?? Or is it that he was only verbal-smart and this is the sort of thing that only makes sense if you've ever been good at linear algebra?? (Such that the language of "only running your clustering algorithm on the subspace of the configuration space spanned by the variables that are relevant to your decisions" would come naturally.) Did I need to write a post explaining just that one point in mathematical detail? (With executable code and a worked example with entropy calculations.)
-
-My dayjob boss made it clear that he was expecting me to have code for my current Jira tickets by noon the next day, so I resigned myself to stay at the office late to finish that.
-
-But I was just in so much (psychological) pain. Or at least—as I noted in one of a series of emails to my posse that night—I felt motivated to type the sentence, "I'm in so much (psychological) pain." I'm never sure how to intepret my own self-reports, because even when I'm really emotionally trashed (crying, shaking, randomly yelling, _&c_.), I think I'm still noticeably _incentivizable_: if someone were to present a credible threat (like slapping me and telling me to snap out of it), then I would be able to calm down: there's some sort of game-theory algorithm in the brain that subjectively feels genuine distress (like crying or sending people too many hysterical emails) but only when it can predict that it will be either rewarded with sympathy or at least tolerated. (Kevin Simler: [tears are a discount on friendship](https://meltingasphalt.com/tears/).)
-
-I [tweeted a Sequences quote](https://twitter.com/zackmdavis/status/1107874587822297089) (the mention of @ESYudkowsky being to attribute credit, I told myself; I figured Yudkowsky had enough followers that he probably wouldn't see a notification):