Screwtape

• ScrewtapeApr 7, 2025, 6:12 PM

  10 points

  4

  in reply to: LVSN’s comment on: The Lizard­man and the Black Hat Bobcat

  I’m going to note I’m having a little trouble parsing your sentences here.

  Strong downvoted for not just saying what you’re really thinking to the person you have a criticism about which is almost definitely wrong.

  I think the thing you’re saying is that you downvoted because you think instead of writing this essay, I should have told a specific person that I think they’re being some kind of jerk (mailing metaphorical bobcats) to a small number of people while being nice to the majority of people. Further, that I’m incorrect about how bad the jerkishness is. Is that close?

  Downvote as you will. But I’m trying to talk about a pattern I’ve noticed across multiple people, and I’m trying to share a tool because I can’t be in every ACX meetup in the world. I want local organizers to be able to notice this faster.

  Also, speaking directly to the person I have a criticism about isn’t always enough. Imagine Bob is at a meetup and is nice to nine other attendees, but punches the tenth in the face in a way only me and the victim can see. The victim leaves, never to return, and I tell Bob not to do that again. Next week, there’s Bob, the nine from last time, and a newcomer. Bob punches the newcomer in the face where only me and the newcomer can see, the newcomer leaves, and I ban Bob. This is the situation I’m talking about in part IV but more-so because I saw the punching myself, and I’m still going to have to explain to the nine regulars why I banned someone who was only ever nice to them.

  (I believe targeting dynamics do happen sometimes—see part V where I at least touch on this - but I also think the basic pattern of “nice to most people but terrible to a few” does happen sometimes.)

• ScrewtapeApr 7, 2025, 5:53 PM

  7 points

  0

  in reply to: Viliam’s comment on: The Lizard­man and the Black Hat Bobcat

  Yep, and also as things scale you just get less information about everyone.

  An random local meetup might fit in one room, sometimes splitting into two rooms so it’s easier to have multiple conversations. I can have line of sight to everyone at once and hear it if voices start getting raised. With meetups in ten cities, I can at least wave at most attendees, and have had a couple hours of conversation with the organizers. With meetups in a hundred cities, I have only demographic guesses about who the attendees are, and it takes time and effort to get to meet the organizers in a way where I’d notice an “I Eat Puppies” tattoo on their face.

  Weirder things can happen, and also you have less bandwidth to notice it.

  Edit: There’s an essay that’s shaped some of my thinking here called Default Blind. I recommend it.

• ScrewtapeApr 7, 2025, 5:40 PM

  4 points

  0

  in reply to: Said Achmiz’s comment on: The Lizard­man and the Black Hat Bobcat

  Somewhat agreed.

  I’m trying to point at something loosely in this vicinity in section V, about hunting in packs—replace “one of them has three good friends” with “one of them paid three people”—where sometimes a bunch of negative reports are happening because someone is making up or deeply exaggerating accusations and routing them to you through different sources. I don’t know that it’s my first assumption; I currently think “Erin is mailing metaphorical bobcats to a small number of people” happens more often than “Frank is coordinating a bunch of fake or exaggerated accusations about Erin.” I do think the latter happens sometimes though.

  Conservation of expected evidence, if I see an Amazon review page with reviews saying “instead of office chair, package contained bobcat” my odds they’re sending bobcats did go up at least somewhat. How much depends on things like how outlandish the actual accusation is and how coordinated the reviews look.

  I strongly agree that adversarial optimization exists and is an important factor in these kinds of social situations. I’ve been thinking about and writing about social conflict a lot lately in large part because the sometimes adversarial nature of it makes this harder in many ways than most parts of running good meetups. Many (most?) problems you can fix and they stay fixed, and if you’re stuck you can ask the people around you. That doesn’t work as well for social conflict.

The Lizard­man and the Black Hat Bobcat

Monty Variations

I’m re­sign­ing as Meetup Czar. What’s next?

• ScrewtapeMar 26, 2025, 8:25 PM

  5 points

  0

  in reply to: Said Achmiz’s comment on: So­cially Grace­ful Degradation

  Huh. That article does not have as much information as I want on how that election process works, but I’ll swap to William The Conqueror as an example. Thanks for pointing it out.

  It’s the second example I’ve had to swap which probably should dock me some kind of points here, though I still feel pretty good about the overall thesis.

• ScrewtapeMar 26, 2025, 8:14 PM

  4 points

  0

  in reply to: Viliam’s comment on: So­cially Grace­ful Degradation

  Even if a skill isn’t as useful if you’re the only one to know it, if the skill is still somewhat useful that can work. I like literacy as an example; crazy good if most people have it, still useful if only you have it, usually obvious pretty quickly if other people don’t have it.

  Individual and group rationality are pretty relevant here. In a sense, one thing I’m pointing at is a way to bootstrap (some) rationality skills from the easier individual domain in to the harder group domain; focus on places where the same skill is relevant in both arenas. It’s also a small argument in favour of following a textbook; mandatory education is one of society’s big shots on putting skills in everyone’s heads and it might not be worth (making numbers up) a 50% boost to this one classroom overall if it means they happen to miss a particular skill that the rest of society is going to expect everyone to have. Still, that side of things is more tentative.

  (One of my favourite questions to ask rationalists is “if you could pick one rationalist skill and make it as common as literacy, what do you pick?”)

• ScrewtapeMar 26, 2025, 8:05 PM

  4 points

  0

  in reply to: simon’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  If I try this again next year I plan to use the exact same text and values on both sides, which hopefully will clear up most of that kind of issue. It doesn’t really fix marginal value, but I’m not sure that’s fatal to this kind of analysis- I can quote a reasonable price for an apple even though my marginal value of apples drops very fast by the time I hit three digits of apples. I could try and fix this by picking things I think people value vaguely the same but then we miss out on catching scope insensitivity.

• ScrewtapeMar 26, 2025, 8:01 PM

  2 points

  0

  in reply to: Garrett Baker’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  11.2% is if I remove the CFAR attendees. 36.8% is if I remove the non-attendees. Possibly this is a backwards way of setting things up but I think it’s right?

  Say I have a general population and I know how many pushups they can do on average (call this Everyone Average), and I remove everyone who goes to the gym and see how many pushups those remaining can do on average (Call this Gym-Removed Average) and then I go back to the general population again this time removing everyone who doesn’t go to the gym (Call this No-Gym-Removed Average.)

  This is a confusing label scheme but I don’t immediately know what the better one is.

  If No-Gym-Removed Average < Gym-Removed Average, then it looks like the gym helps.

  (Totally possible I’m screwing something up here still)

Bos­ton – ACX Mee­tups Every­where Spring 2025

Berkeley – ACX Mee­tups Every­where Spring 2025

• ScrewtapeMar 22, 2025, 1:51 AM

  2 points

  0

  in reply to: Ben’s comment on: So­cially Grace­ful Degradation

  No, I think I’m actually just wrong here and River is correct. I don’t know how I wound up with the clockwise rule in my head but I just checked the new driver’s pamphlet and it’s first to the intersection. Updated.

So­cially Grace­ful Degradation

• ScrewtapeMar 18, 2025, 9:12 PM

  2 points

  0

  in reply to: Alice Blair’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  but predicted that it was instead about sensitivity to subtle changes in the wording of questions.

  If I try this again next year I’m inclined to keep the wording the same instead of trying to be subtle.

  Regarding the dutch book numbers: it seems like, for each of the individual-question presentations of that data, you removed the outliers. When performing the dutch book calculations, however, it seems like you keep the outliers in.

  Yep. Well, in the individual reports I reported the version with the outliers, and then sometimes did another pass without outliers. I kept all of the entries that answered all the questions for the dutch book calculations, even if they were outliers. I think this is the correct move: if someone’s valuations are wild outliers from everyone elses but in a way that multiplies out and gets them back to a 1:1 ratio, then being an outlier isn’t a problem.

  (Imagine someone who values a laptop at one million bikes, and a bike at equal to one car, and a car at one millionth of a of a laptop. They’re almost certainly a wild outlier, and I’m confused as heck, but they are consistent in their values!)

• ScrewtapeMar 18, 2025, 8:54 PM

  2 points

  0

  in reply to: Garrett Baker’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  Hrm. I guess what would be helpful here would be a sense of the range; the average briers floated around .20 to .23, and I don’t have a sense of whether that’s a tight clustering with a bit of noise or a meaningful difference. To use running a mile as a comparison, differences of seconds mostly aren’t important (except at high levels) but differences of minutes are, right?

• ScrewtapeMar 18, 2025, 8:45 PM

  2 points

  0

  in reply to: mako yass’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  If Other is larger than I expect, I think of that as a reason to try and figure out what the parts of Other are. Amusingly enough for the question, I’m optimistic about solving this by letting people do more free response and having an LLM sift through the responses.

• ScrewtapeMar 15, 2025, 10:52 PM

  3 points

  0

  in reply to: jenn’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  Thank you! I felt quite clever setting it up.

• Screwtape 15 Mar 2025 21:38 UTC

  2 points

  0

  in reply to: SarahNibs’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  Yeah, I should probably add a bit at the start or end of that section that everything in it is potentially selection effect. I don’t know how to look at the thing I’m curious about without that.

  Thinking out loud: If you get a random selection of people from the Pushup Club and count how many pushups they can do, then do the same for general population, the difference could be selection effect. People who like doing pushups are more likely to go to pushup club in the first place, and more likely to stick with it. But I can’t realistically pay a bunch of Mechanical Turkers to hang out on LessWrong for six years and watch what happens. Presumably there’s some approach actual scientists have here, but I don’t know what it is. Suggestions welcome.

  In the mean time I’m going to add a bit towards the start of the section warning of potential selection effects.

• Screwtape 15 Mar 2025 21:16 UTC

  2 points

  0

  in reply to: Sherrinford’s comment on: 2024 Unoffi­cial LessWrong Sur­vey Results

  No, I think that’s correct.

  There’s 107 people who answered above 200, 21 who answered exactly 200, and 113 people who answered below 200. The second quartile (aka the median) is 200. But nobody guessed a negative number, so the people who guessed low aren’t pulling the mean down that much. Meanwhile 33 people guessed 1000 or higher, and they can yank the mean a lot without doing that much to the median. If you’re asking people to generate numbers, you tend to get whole number quartiles because nobody guesses there’s 100.5 stations.

  Imagine a the set [1,1,1,2,2,2,2,100,100]. The average is ~23.444, but the median is 2.

  Or have I misunderstood the thing that you think needs to be corrected?