In a recent in-person discussion about moderation, habryka brought up the claim that people do way better work when they can build up social contexts around doing that work. Rather than have one moderator flying solo, have multiple moderators who can build up a complicated theory of moderation and work out decision trees and mid-level considerations and give durable labels to them. At the time, I didn't buy it, but after consideration decided that I did buy a related thing.

I buy that people organize their brains through conversation, but I worry that the "20-hour team conversations" make things more opaque instead of clearer. That is, if you have a social context of chefs developing lots of shared chef concepts, this makes food better instead of worse (because customers don't need to understand chef-speak but can taste the improvement), but if you have a social context of judges and lawyers developing lots of shared legal concepts, this makes the law more opaque instead of clearer (because citizens do need to understand legalese in order to interpret the law).

One of the things that was surprisingly helpful was to go on a walk with a friend, who hadn't thought much about moderation issues (but was familiar with the community and the Sequences and so on) and talk through some of the considerations; it both allowed me to organize thoughts and had the constraint of "I can't import any dependencies that aren't actually shared" and prevented me from accidentally doing so.

Relatedly, I think discussions on LW about moderation are probably net good, because they put moderators in touch with what the users actually care about, and responsiveness is probably more important for durable trust than initially having the correct position, so long as the presumption of the moderators (and the users?) is still something like "make sure the obviously-good moderation can get done."

Continuing the thread:

I was talking specifically about about moderation and matters of LW policy and norms. I think I see the value of being able to freely express our thoughts without concern for consensus, but I now think there are factors whose importance competes the value of being able to talk freely online.

The update (which I kind of contributed to the team, but which I think Ray first crystallized and propagated) occurred in the wake of a number of lengthy (20+ hour?) conversations we were having with people on- and offline a few month ago about moderation stuff. I think you were on Sabbatical then, Ben, which might be why you didn't share this update as strongly as me or Ray. The update was that notwithstanding caveats, people remained confused (and very concerned) in the wake of us "thinking aloud".

To quickly sketch out some factors I don't think we can ignore and might warrant be more carefully what we say:

• The five of us are extremely powerful custodians of a public commons that many people are heavily invested and care deeply about . By "extremely powerful", I mean that structurally we are able to govern it and enact our decisions without requiring any kind of affirmative assent from the community. The community has informal means of complaint (which are quite powerful), but this isn't obvious to everyone and I think that many people fear that if we start taking things in the wrong direction, there goes LW.
  • In light of that, I think it's a reasonable reaction for people, in the face of hearing a LW moderator espouse a view about how LW might go (one they like or fear) to make an update that LW might actually enact such a view and therefore have a reaction of pleasure, fear, or panic in response to hearing it; or at the very least update their models of what might happen.
  • 1) It's non-standard or members of an organization to freely express individual ideas of policy, 2) It's cognitively difficult to keep track of five (six if you include Vaniver) models of moderation when trying to model what LW policy is. I don't blame people if they end up lumping things we individually say together, losing track, and getting confused about what's going to happen with their public commons.
• In a world where any opinion we express gets taken seriously by people as expression of what LessWrong is, we're subject to unilateralist's curse. A and B of us might hold back from expressing an individual opinion we don't want people to mistakenly to attribute as proper LW policy, but then C thinks it's fine and does so, at which A and B feel obligated to swoop in and correct the record, leading to . . .
• Appearances of breakdown of internal team communication. This is gnarly consideration, but if people are to trust as a team, they need to believe we're capable of resolving disagreements and deciding joint policy (especially for high-level vision stuff, culture, and mod policy). It can be both good and risky for us to debating this publicly. Good because it allows others to participate in our discussion, bad if it looks like we failed to communicate internally and are now duking it out publicly, trying to correct each other, etc. I think it's fair for people to update a big negatively depending on how that happens.

My analogy for punchiness:

Suppose you're a farmer who fled your old kingdom which was ruled by an incompetent tyrant to a new kingdom ruled by the council of "five fairly-wise men." You believe the rulers are well intentioned, but also they're not elected and get to enact their wishes almost immediately without restraint. 

When you hear one of the men of the council opining in the town square about grain taxation and water quota policies that would seriously affect you, you have reason to get worried that this might soon become reality– even if the council elder proclaims "this is just my individual view, I'm just thinking aloud" It's worse if you only stumble upon the conversation part way through and missed his disclaimers that were three comments earlier, or you forgot in the midst of the lengthy rant.