You can be a moderate by believing only moderate things. Or you can be a moderate by adopting moderate strategies. These are not necessarily the same thing.
How does one end up with moderate beliefs without relying on moderate strategies? (In less pressured fields, I could imagine this happening as a matter of course, but I am surprised if someone follows a reality-hugging strategy in AI and ends up believing 'moderate things'.)
I also think this… I think? I guess it depends on what exactly you mean by “the polite and detailed version”.
Then perhaps I misunderstand what you mean by "corrective authority"? It seems to me like "read the Sequences" is an example of "apply such measures as will make the people in your system alter their behavior, to conform to relevant optimality criteria". But then I find it difficult to square with:
attempting to behave as a “corrective authority”, in the context of a forum like this, is weird and bad.
Perhaps the difference is between "read the sequences" and "if you keep posting low-quality comments, we will ban you, and this part of the sequences explains the particular mistake you made here"? Or perhaps the difference is between the centralized moderator decision-making ("this comment is bad because Alice says so and her comments have a fancy border") and decentralized opinion-aggregation and norm enforcement ("this comment is bad because its net karma is negative")?
There is a different way to make things coherent, of course, which is that as part of the transition to LW 2.0 the mod team attempted to shift the culture, which involved shifting the optimality criteria, and the objection to us being corrective authorities in this way is not an objection to corrective authority as a method but instead an objection to our target. Which, that's fair and not a surprise, but also it seems like the correct response to that sort of difference is for us to shake hands and have different websites with different target audiences (who are drawn to different targets). Otherwise we'll just be locked in conflict forever (as happens when two control systems are trying to set the same variable to different reference values) and this doesn't seem like a productive conflict to me. (I do think we've written about culture and Zack has written about culture downstream of this disagreement in a way that feels more productive than the moderation discussions about specific cases, but this feels way worse than, say, artists jockeying for status by creating new pieces of art.)
there just is not any such thing as some single “true rejection”.
I think this is correct, in that many decisions are made by aggregating many factors, and it's only rarely going to be the case that a single factor (rather than a combination of factors) will be decisive.
(I do note this is a situation where both of us 'disagree with the Sequences' by having a better, more nuanced view, while presumably retaining the insight that sometimes decisive factors are unspeakable, and so discussions that purport to be about relevant information exchange sometimes aren't.)
Otherwise, I must say that I do not at all appreciate you talking as if the decision isn’t final, when in fact it is.
Fair. I think it is challenging to express the position of "New information could persuade me, but I don't expect to come across new information of sufficient strength to persuade me."
(On the related stakeholders point: I agree that it is often vague, but in this specific case I'm on the board that can decide to fire habryka, and one of the people who is consulted about decisions like this before they're made. I suspect that in the counterfactual where I left the mod team at the start of 2.0, you would have been banned several years earlier. This is, like, a weird paragraph to write without the context of the previous paragraph; I was in fact convinced this time around, and it is correspondingly challenging to convince me back the other direction, and it seems cruel to create false hope, and difficult to quantitatively express how much real hope there is.)
an offer which, as I expect you know, is not an idle one, when coming from me!
Indeed; I have appreciated a lot of the work that you've done over the years and am grateful for it.
It is also honest about the relative “consensus value” of the opposing views
Something about the "consensus value" phrasing feels off to me, but I can't immediately propose a superior replacement. That is, it would be one thing if just Oli disagreed with you about moderation and another different thing if "the whole mod team disagrees with Said about moderation". The mods don't all agree with each other--and it took us years to reach sufficient agreement on this--but I do think this is less like "two people disagree" and more like "two cultures are clashing".
That said, I do think I see the thing that I could have noticed if I were more alert, which is that I already had the view that we were optimizing for different targets, and making that the headline has more shared-reality nature to it. Like, I think the following framing is different from yours but hopefully still seems valid to you:
Since the beginning of LW 2.0, and the mod team's attempts to move LessWrong's culture in a direction that we thought would be more productive for our broader goals, we have been disagreeing with Said about which cultural elements are features and which are bugs. We think this is downstream of differences of preference, principle, and experience. Because of Said's many positive qualities and many beneficial contributions to the site, the mod team has spent quite a bit of effort on attempting to persuade him to move in our direction, and I personally have spent about a hundred hours a year on moderating Said and his influence on LW's culture. Today I am declaring defeat on the goal of getting Said to not shape LessWrong's culture in directions I think are bad for our goals, and am giving him a 3 year ban.
what do the other mods think of it?
I asked in the sunshines channel on the LW slack and people there said that they were voting comments based on quality as a comment, and while one is downvoting many of your comments on the page overall, was not downvoting the majority of the comments in this thread.
There are more 10-strength users than just the mods; it may be the case that enough of them are downvoting comments that are at positive karma but leaving the -8 comments alone, which results in no one person downvoting more than a few comments in the thread, but the comments being underwater as a whole. But if there is a single mod who is trying to make this thread not happen, they're not telling me (which seems worth doing because it would affect my behavior more than the downvoting would). [Edit: the person who did the database query clarified, and I now think that the votes are primarily coming from mods.]
I made the classic mistake of 'asking two questions together' and so primarily got responses on voting behavior and not what they think of the project, but I would (from their other writing) guess they are mostly out of hope about it.
My model of Said would have been offended by being asked to take on a Jester role as a condition of staying on LessWrong, but perhaps he would have been interested?
I do think the background culture that we're in is one that doesn't really have this role anymore, because comedians are sometimes major public figures whose opinions are treated with respect; people don't dismiss what they say just because they're frivolous.
I think "you can't get the coffee if you're dead" goes thru just as well for "you can't get the coffee if you're preserved in a vault"; do you not?
And you want me to explain why these things are bad?
Yes. Part of this is because my long experience is that sometimes our sense of communication or our preferences for norms have flipped signs. If you think something is bad, that's moderate but not strong evidence that I think it's bad, and we might be able to jump straight to our disagreement by trying to ground out in principles. I think in several previous threads I wish I had focused less on the leaves and more on the roots, and here was trying to focus on roots.
If so, then—sure, I don’t in principle object to such exercises—on the contrary, I often find them to be useful—but why do this here, now, about these specific things? Why ask me, in particular?
...
On the other hand, if what you’re saying is that you disagree that the aforementioned things are bad, then… I guess I’m not sure how to respond to that, or what the point would even be…
I mean, I am genuinely uncertain about several parts of this! I think that the audience might also be uncertain, and stating things clearly might help settle them (one way or the other). I think there is value in clear statements of differences of opinion (like that you have a low opinion of interpretative labor and I have a high opinion of it), and sometimes we can ground those opinions (like by following many conversations and tracking outcomes).
Like, I understand 'tendentious' to be a pejorative word, but I think the underlying facts of the word are actually appropriate for this situation. That doesn't mean it's generically good, just that criticizing it here seems inappropriate to me. Should we not invite controversy on ban announcements? Should we not explain the point of view that leads us to make the moderation decisions we make?
But perhaps you mean something narrower. If the charge is more "this is problem only a few users have, but unfortunately one of them is an admin, and thus it is the site rule"--well, we can figure out whether or not that's the case, but I don't actually think that's a problem with the first paragraph, and I think it can be pointed at more cleanly.
Well, the “sneaking in connotations” bit is a link to a Sequence post (titled, oddly enough, “Sneaking in Connotations”). I don’t think that I can explain the problem there any better than Eliezer did.
As it happens, I reread that post thru your link. I thought that it didn't quite apply to this situation; I didn't see how habryka was implying things about you thru an argument via definition, rather than directly stating his view (and then attempting to back it up later in the post). I thought Frame Control would've been a better link for your complaint here (and reread our discussion of it to see whether or not I thought anything had changed since then).
The other stuff really seems like it’s either self-explanatory or can be answered with a dictionary lookup (e.g., “begging the question”).
I also didn't quite buy that "begging the question" applied to the first paragraph. (For the audience, this is an argument that smuggles in its conclusion as a premise.) I understood that paragraph to be the conclusion of habryka's argument, not the premise.
Overall, my impression was--desperation, or scrambling for anything that might stick? Like, I think it fits as a criticism of any post that states its conclusion and then steps thru the argument for that conclusion, instead of essaying out from a solid premise and discovering where it takes you. I think both styles have their virtues, and think the conclusion-first style is fine for posts about bans (I've used it for that before), and so I don't find that criticism persuasive. (Like, it's bad to write your bottom line and then construct the argument, but it's not bad to construct an argument and then edit your introduction to include your conclusion!)
But maybe I missed the thing you're trying to convey, since we often infer different things from the same text and attend to different parts of a situation. I tried to jump us to the inferences and the salient features, and quite possibly that's not the best path to mutual understanding.
What’s the purpose of this line of inquiry?
There are several. The overarching goal is that I want LessWrong's contribution to global cognition to be beneficial. As a subgoal to that, I want LessWrong's mod team to behave with integrity and skill. As subgoals to that, I'm trying to figure out whether there were different ways of presenting these ideas that would have either worked better in this post, or worked better in our discussions over the years at grounding out our disagreement; I'm also interested in figuring out if you're right and we're wrong!
Related to the last subgoal, I think your typology of selective/corrective/structural is useful to think about. I view us as applying all three--we screen new users (a much more demanding task now that LLMs are directing people to post on LessWrong), we give warnings and feedback and invest some in rationality training projects, and we think about the karma system and UI changes and various programs and projects that can cause more of what we want to see in the world. I don't think behaving as a corrective authority is weird and bad; I think the polite and detailed version of "read the sequences" is good.
But more narrowly--looking at this conversational chain--you made a criticism of habryka's post, and I tried to take it seriously. Does it matter that the post expresses or promotes a particular point of view? Does it matter that it's controversial? What would it look like to fix the problems in the first paragraph? I left comments on an earlier draft of this post, and I tried to apply a framework like "how would the other guy describe this?", and I missed those problems in the first paragraph. Tsuyoku Naritai.
[I think that you deserve me giving this a real try, and that the other mods deserve me attempting to get to ground on something with you where we start off with a real disagreement, or where I don't understand your position.]
I understood us to be discussing a thing that Habryka wrote in the post. If the thing he wrote involves power relations, or connotations about power relations, then how can we not be discussing power relations…?
Reductionism--the idea that things are made out of parts. We can focus on different parts of it at different times. To me this also relates to the idea of True Rejections. If what you are objecting to is that habryka is banning you and that he's the mod and you aren't, then--I feel sympathy for you, but there's really not much to discuss. I think there is a lot to discuss about whether or not it's right for LW to ban you, because I am pretty invested in pushing LW to do the right thing. And that one is not a power relations question, and seems like one that we can discuss without power relations.
Yes, even if we construct airtight arguments, habryka might still ignore them and go through with the ban anyway. Yes, some people will reflexively support the mods because they like the website existing and want to subsidize working on it. But some people are watching and thinking and deciding how to relate to LW moving forward based on how these arguments shake out. That is...
But of course his standards can’t be controversial, because he’s the admin. If someone disagrees with his standards—irrelevant; he doesn’t have to care.
I think there are meaningful stakeholders whose disapproval would sink habryka's ability to run LessWrong, and I think attempting to run LessWrong in an unethical or sloppy way would lead to the potential benefits of the site turning to ash.
(I also think this is a nonstandard usage of 'controversial'. It just means 'giving rise to public disagreement', which moderation decisions and proposed norms and standards often do. Like, you're controverting it right now!)
Returning to true rejections--suppose a fundamental issue here is that you have one vision for LW, where there's no corrective authority, and we have a different vision for LW, where there is corrective authority. Then I think either we find out why we want those things and identify cruxes and try to learn more about the science of communication and moderation so that we can better achieve our shared goals, or we decide that our goals are sufficiently in conflict that we should pursue them separately. And, like, the value I see in habryka's offer to edit in your text to the post is that you can make your pitch for your vision, and maybe people who prefer that vision will follow you to Data Secrets Lox, and the more clarity we can reach the more informative that pitch can be.
FWIW, this guess is so far removed from being right that I have trouble even imagining how you could have generated it. (Yet another in a very long series of examples of why “interpretive labor” is bad, and trying to guess what one’s interlocutor thinks when you already know that you don’t understand their view is pointless.)
What do you think, then? Why are those things bad and why is objecting to them good?
If you can't answer those questions, then I'm not sure what arguments about propriety we could have. If we are to design functional site norms, we should be guided by goals, not merely following traditions.
(The point of interpretive labor, according to me, is to help defeat the Illusion of Transparency. If I read your perfectly clear sentence and returned back a gross misunderstanding--well, then a communication breakdown happened somewhere. By looking at what landed for me, we have a stacktrace of sorts for working backwards and figuring out what should have been said to transmit understanding.)
Of course it’s by his lights. How else could it be?
Some people realize that their position is a personal one; others assume that their position is standard or typical. Such phrases are often useful as evidence that the person realizes that fact; of course, since they can be easily copied, they are only weak evidence. "Strawberry is a better flavor, according to me" is a different sentence from "Strawberry is a better flavor", and those two are yet again different from "Four is larger than two." Adding 'according to me' to the last option would be a joke.
I think a frequent source of conflict has been differing judgments on what is usual and what is unusual, or what is normal and what is abnormal.
The formulation you offer connotes a scenario where two parties enter into discussions and/or negotiations as equals, without presupposing that their own view is necessarily correct or that no compromises will need to be made, etc. But of course nothing remotely like that was the case. (The power relation in this case has always been massively asymmetric, for one thing.)
I understood us not to be discussing power relations (was anyone ever confused about who was the admin of LessWrong?) but something more like legitimacy relations (what should be the rules of LessWrong?). You've been here longer; you might know the Sequences better; you might have more insight into the true spirit of rationality than habryka. In order to adjudicate that, we consult arguments and reasons and experience, not the database.
Using the lens of power relations, your previous complaint ("This phrasing assumes") seems nonsensical to me; of course the mod would talk about educating the problem user, of whether they understand and learn the models and behaviors as handed down from on high.
Here I would like to take a step outward and complain about what I perceive as a misstep in the conversational dance. Having criticized habryka's paragraph, you describe its flaws and went so far as to propose a replacement:
I have had disagreements with Said; we have discussed, debated, argued; I remain convinced of my view’s correctness.
My replacement differs from yours. But I claim this criticism of my replacement (that it connotes a discussion of equals) applies just as readily to yours, if not more readily because my version includes the ban. (A more fair comparison probably ends at 'on that goal' and drops the last phrase.) If not, it is for minor variations of style and I suspect any operationalization we come up with for measuring the difference (polling Turkers or LLMs or whatever) will identify differences between their connotations as minor (say, a split more even than 66-34 on which connotes more even power relations).
Here my thoughts turn to the story in The Crackpot Offer, and the lesson of looking for counterarguments to your own counterarguments.
I thought people exchange between adjacent tribes was common, which makes me suspect that the amount it time it takes to transfer innovations is actually not that long?