[ Question ]

LW authors: How many clusters of norms do you (personally) want?

by Raemon 3mo7th Jul 201940 comments

40


note: this post is about gathering data for "what might the archipelago model look like if implemented on LW", which is s separate question from "should the archipelago model be implemented on LW?"

Periodically, I've argued for applying Archipelago-style norms to LessWrong (i.e. give people the tools to establish different norms and culture, then let them experiment however they want, and let the best cultures attract participants)

One cost of this approach is that it's much harder to keep track of which space you're in, and what the rules are. This could be disorienting.

There are various ways to streamline that process. For example, we could (and I expect will, at least on desktop computers) make it so that as soon as you start typing a comment, the moderation norms for the relevant post appear next to your comment box, so it's much easier to see what norms you're expected to follow.

But if there were *hundreds* of different types of norms, this might still get a bit bewildering (in particular if each set of norms had a lot of nuance to it), and place too high a burden on commenters.

One thing I'm wondering is how many different normsets there actually are demand for, among authors. (I do think it's important to ground this out in "what authors want" rather than "what commenters want" since the authors are doing the bulk of the work, and the conversations won't happen at all without them, with the caveat that I think it's a fine outcome for an author to write a post with one set of comment-norms, and someone who prefers other norms to write up a post titled "Discussing Bob's Post X" that sets different norms).

Knobs that I could imagine an author wanting to turn include:

  • Do you want feedback on writing style?
    • (possibly with different answers for typos, high level structure, etc)
  • Is there a particular conversation you're trying to have? Is there a product you're trying to build? Or are your mostly just throwing your ideas out into the wild and see what happens?
  • Should commenters relate to your post as if it's in a playful/creative/exploratory stage, or more of a finished product that you want judged professionally (or somewhere else on a spectrum?
  • Do you prefer that you (and each commenter) be responsible for justifying their ideas, or do you prefer commenters who are helping each other and you to figure out whether an idea is good?
  • How much do you expect people to own their own emotional state?
  • Is this intended to be a 101-discussion (or, a "201" discussion where participants have read the sequences and major-upvoted-posts-since-then, but not necessarily much else), or are you knee deep in a particular frame and you aren't interested in commenters who don't share that frame?

It'd be convenient if this turned out to compress into 2-4 major sets of norms (which could then be made clearly visually distinct)

As an author, or as a person who thinks they'd actively be motivated to write posts akin to "Discussing Bob's Ideas X through Normset Z", what are the commenting norms you'd want to have?

40

New Answer
Ask Related Question
New Comment
Write here. Select text for formatting options.
We support LaTeX: Cmd-4 for inline, Cmd-M for block-level (Ctrl on Windows).
You can switch between rich text and markdown in your user settings.

6 Answers

I can think of a few natural clusters that might not satisfy everyone's every preference, but I think come close to at least letting you know which kind of space you are in. I'd be willing to compromise on my own norms if there was a nearby enough cluster that gave me 80% or better of what I want. These are based loosely on what I perceive to be the clusters people have asked for or directly expressed in commenting guidelines here and elsewhere.

Suggested norm clusters:

  • Debate/Argue/Combat - This is the cluster for typical ways of arguing things out, where one person says something, another who disagrees tries to refute it, and there's a back and forth of disagreeing. I think of this as the LW 1.0 cluster: very smart people calling you out on things so we can collectively do better by highlighting mistakes and correcting them. This cluster favors direct communication with minimal concern for managing the feelings of others.
  • Dialectic/Inquire/Cooperate - This is the cluster for digging into things by applying maximum charity, trying to understand what someone else is saying, and helping to improve their arguments rather than tearing them down when you notice a mistake. This looks different from the first cluster in that, for example, when you think someone is wrong you don't tell them they're wrong bluntly, you would ask them "hey, this seems to contradict this other thing and I can't figure out how they work together; what do you think?". You could also skewedly think of this as the NVC cluster, where each person is making space to prevent anyone getting defensive or feeling attacked.
  • Discuss/Share/Chatter - This cluster doesn't really existing on LW now because it's not primarily optimized for truth seeking, but things like this exist all over the internet. This is the "let's talk like nice, civilized people at a dinner party" cluster. No one says anything mean, everyone tries to help each other save face, and if you have something to say you bring it in the spirit of sharing only with no intention that you change anyone's beliefs (other than maybe that they now believe that you believe something).

Maybe there could be an "asshole" cluster opposite the third "nice" cluster, but I'm not sure that serves any serious purpose here on LW; people already have plenty of places to talk that way to each other and it seems anti-conducive to the general LW project, whereas the "nice" cluster could serve a purpose on some kinds of posts or for some kinds of authors that would not bring down the general quality of LW (it would just be neutral in terms of truth seeking, I think).

I think I don't have any strong object-level preferences along these lines, and if norm clusters develop, will probably end up copying/adopting whatever norm cluster that seems to produce the most vibrant, highest quality discussions.

One thing I would really like though, is a chance to experiment with this idea to see what effects it has on discussions (hopefully positive ones), and would definitely enable it for myself if it was an option that authors could choose for their comment sections.

I would post much more on lesswrong if there was a 'no nitpicking' norm available.

(re-posted as a top level comment at Ray's request)

I ideally want ~one set of norms, even if it doesn't agree with my preferences on all points. The cost of illegibility and decentralization is not in my view worth the benefit of being able to fiddle with everything.

I do think it's possible for one set of norms to basically include all the stuff you mentioned as different knobs, though.

Answering for myself on my own posts:

  • I almost always have a product I'm trying to build, and a particular frame for the conversation that I'm most interested
    • I think people showing up with other frames that explore different directions is almost always good. All else being equal more comments is better even if they aren't quite the frame I had.
    • People who make demands on my time that aren't related to the conversation I wanted to have are at least somewhat negative for me, and on the margin make me want to post somewhere else or not at all.
  • I usually prefer getting feedback on writing style/quality
  • Ideas vary in terms of "early stage" vs "late stage", and generally in the early stage I care more about people understanding my frame, and in the later stages I think it makes sense for people to critique it in more detail, from a variety of frames.
  • My sense of "who is responsible for emotional state" is sort of complicated and maybe beyond scope for this comment. But also, I don't think I usually run into this problem on my own posts, in part because I try to design posts such that disagreements play out in a way I consider productive.

I generally prefer norms that look like sparring - anything that's relevant is fair game, anything on the boundary of personal attack is fair game so long as you can make the case for its relevance.

Personal preferences aside, the biggest norm problem I've encountered is when people make an assertion based on priors that are taboo to discuss but you can't make a solid counterargument without addressing them.